docs: add Space Bunny to V2 console docs (#50940 )

fix(stats): attribute Hy4 preview traffic to Tencent (#50935 )
Co-authored-by: MrMushrooooom <19261047+MrMushrooooom@users.noreply.github.com>
2026-09-23 17:17:43 +00:00 · 2026-09-24 01:03:39 +08:00 · 2026-09-24 00:56:42 +08:00 · 2026-09-23 12:36:10 -04:00 · 2026-09-23 11:33:49 -05:00 · 2026-09-23 11:27:33 -05:00
204 changed files with 11862 additions and 895 deletions
@@ -32,7 +32,7 @@
    },
    "packages/ai": {
      "name": "@opencode/ai",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@aws-sdk/credential-providers": "3.1057.0",
        "@opencode/schema": "workspace:*",
@@ -54,7 +54,7 @@
    },
    "packages/app": {
      "name": "@opencode/app",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@corvu/drawer": "catalog:",
        "@dnd-kit/abstract": "0.5.0",
@@ -112,8 +112,9 @@
    },
    "packages/cli": {
      "name": "@opencode/cli",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "bin": {
+        "opencode": "./bin/opencode.cjs",
        "opencode2": "./bin/opencode2.cjs",
      },
      "dependencies": {
@@ -176,7 +177,7 @@
    },
    "packages/client": {
      "name": "@opencode/client",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@opencode/protocol": "workspace:*",
        "@opencode/schema": "workspace:*",
@@ -202,7 +203,7 @@
    },
    "packages/codemode": {
      "name": "@opencode/codemode",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "acorn": "8.15.0",
        "effect": "catalog:",
@@ -215,7 +216,7 @@
    },
    "packages/console/app": {
      "name": "@opencode/console-app",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@cloudflare/vite-plugin": "1.15.2",
        "@ibm/plex": "6.4.1",
@@ -251,7 +252,7 @@
    },
    "packages/console/core": {
      "name": "@opencode/console-core",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@aws-sdk/client-sts": "3.782.0",
        "@jsx-email/render": "1.1.1",
@@ -278,7 +279,7 @@
    },
    "packages/console/function": {
      "name": "@opencode/console-function",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@openauthjs/openauth": "0.0.0-20250322224806",
        "@opencode/console-core": "workspace:*",
@@ -295,7 +296,7 @@
    },
    "packages/console/mail": {
      "name": "@opencode/console-mail",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@jsx-email/all": "2.2.3",
        "@jsx-email/cli": "1.4.3",
@@ -319,7 +320,7 @@
    },
    "packages/console/support": {
      "name": "@opencode/console-support",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@cloudflare/vite-plugin": "1.15.2",
        "@opencode/console-core": "workspace:*",
@@ -339,7 +340,7 @@
    },
    "packages/core": {
      "name": "@opencode/core",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@ai-sdk/cohere": "3.0.27",
        "@ai-sdk/gateway": "3.0.104",
@@ -407,7 +408,7 @@
    },
    "packages/desktop": {
      "name": "@opencode/desktop",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@zip.js/zip.js": "2.7.62",
        "electron-context-menu": "5.0.0",
@@ -456,7 +457,7 @@
    },
    "packages/enterprise": {
      "name": "@opencode/enterprise",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@hono/standard-validator": "catalog:",
        "@opencode-ai/sdk": "1.18.21",
@@ -493,7 +494,7 @@
    },
    "packages/function": {
      "name": "@opencode/function",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@octokit/auth-app": "8.0.1",
        "@octokit/rest": "catalog:",
@@ -509,7 +510,7 @@
    },
    "packages/http-recorder": {
      "name": "@opencode/http-recorder",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@effect/platform-node-shared": "4.0.0-rc.112",
      },
@@ -528,7 +529,7 @@
    },
    "packages/httpapi-codegen": {
      "name": "@opencode/httpapi-codegen",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "effect": "catalog:",
        "prettier": "3.6.2",
@@ -541,7 +542,7 @@
    },
    "packages/latex": {
      "name": "@opencode/latex",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@opencode/plugin": "workspace:*",
        "@opentui/core": "catalog:",
@@ -555,7 +556,7 @@
    },
    "packages/merman": {
      "name": "@opencode/merman",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@opencode/plugin": "workspace:*",
        "@opentui/core": "catalog:",
@@ -570,7 +571,7 @@
    },
    "packages/plugin": {
      "name": "@opencode/plugin",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@ai-sdk/provider": "3.0.8",
        "@opencode/ai": "workspace:*",
@@ -609,7 +610,7 @@
    },
    "packages/plugin-browser": {
      "name": "@opencode/plugin-browser",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@opencode/plugin": "workspace:*",
        "@opencode/schema": "workspace:*",
@@ -639,7 +640,7 @@
    },
    "packages/protocol": {
      "name": "@opencode/protocol",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@opencode/schema": "workspace:*",
        "effect": "catalog:",
@@ -654,7 +655,7 @@
    },
    "packages/schema": {
      "name": "@opencode/schema",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@standard-schema/spec": "catalog:",
        "effect": "catalog:",
@@ -678,7 +679,7 @@
    },
    "packages/sdk": {
      "name": "@opencode/sdk",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@opencode/client": "workspace:*",
        "@opencode/core": "workspace:*",
@@ -699,7 +700,7 @@
    },
    "packages/server": {
      "name": "@opencode/server",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@effect/platform-node": "catalog:",
        "@effect/platform-node-shared": "catalog:",
@@ -721,7 +722,7 @@
    },
    "packages/session-ui": {
      "name": "@opencode/session-ui",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@kobalte/core": "catalog:",
        "@opencode/client": "workspace:*",
@@ -756,7 +757,7 @@
    },
    "packages/simulation": {
      "name": "@opencode/simulation",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@opencode/ai": "workspace:*",
        "@opencode/core": "workspace:*",
@@ -776,7 +777,7 @@
    },
    "packages/stats/app": {
      "name": "@opencode/stats-app",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@ibm/plex": "6.4.1",
        "@kobalte/core": "catalog:",
@@ -810,7 +811,7 @@
    },
    "packages/stats/core": {
      "name": "@opencode/stats-core",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@aws-sdk/client-athena": "3.933.0",
        "@planetscale/database": "1.19.0",
@@ -829,7 +830,7 @@
    },
    "packages/stats/server": {
      "name": "@opencode/stats-server",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@aws-sdk/client-firehose": "3.933.0",
        "@effect/platform-node": "catalog:",
@@ -875,7 +876,7 @@
    },
    "packages/theme": {
      "name": "@opencode/theme",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@opentui/core": "catalog:",
        "effect": "catalog:",
@@ -889,7 +890,7 @@
    },
    "packages/tui": {
      "name": "@opencode/tui",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@opencode/client": "workspace:*",
        "@opencode/core": "workspace:*",
@@ -924,7 +925,7 @@
    },
    "packages/ui": {
      "name": "@opencode/ui",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@kobalte/core": "catalog:",
        "@pierre/diffs": "catalog:",
@@ -959,7 +960,7 @@
    },
    "packages/util": {
      "name": "@opencode/util",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@effect/opentelemetry": "catalog:",
        "@effect/platform-node": "catalog:",
@@ -967,10 +968,14 @@
        "@npmcli/arborist": "catalog:",
        "@npmcli/config": "10.8.1",
        "@opentelemetry/api": "1.9.0",
-        "@opentelemetry/context-async-hooks": "2.6.1",
-        "@opentelemetry/exporter-trace-otlp-http": "0.214.0",
-        "@opentelemetry/sdk-trace-base": "2.6.1",
-        "@opentelemetry/sdk-trace-node": "2.6.1",
+        "@opentelemetry/api-logs": "0.219.0",
+        "@opentelemetry/context-async-hooks": "2.8.0",
+        "@opentelemetry/exporter-trace-otlp-http": "0.219.0",
+        "@opentelemetry/resources": "2.8.0",
+        "@opentelemetry/sdk-logs": "0.219.0",
+        "@opentelemetry/sdk-metrics": "2.8.0",
+        "@opentelemetry/sdk-trace-base": "2.8.0",
+        "@opentelemetry/sdk-trace-node": "2.8.0",
        "cross-spawn": "catalog:",
        "effect": "catalog:",
        "glob": "13.0.5",
@@ -992,7 +997,7 @@
    },
    "packages/web": {
      "name": "@opencode/web",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "@astrojs/cloudflare": "12.6.3",
        "@astrojs/markdown-remark": "6.3.1",
@@ -1033,7 +1038,7 @@
    },
    "services/update": {
      "name": "@opencode/update",
-      "version": "2.0.14",
+      "version": "2.0.15",
      "dependencies": {
        "jose": "6.0.11",
        "semver": "catalog:",
@@ -1068,7 +1073,6 @@
  "trustedDependencies": [
    "electron",
    "esbuild",
-    "protobufjs",
  ],
  "patchedDependencies": {
    "@pierre/trees@1.0.0-beta.4": "patches/@pierre%2Ftrees@1.0.0-beta.4.patch",
@@ -2220,31 +2224,31 @@

    "@opentelemetry/api": ["@opentelemetry/api@1.9.0", "", {}, "sha512-3giAOQvZiH5F9bMlMiv8+GSPMeqg0dbaeo58/0SlA9sxSqZhnUtxzX9/2FzyhS9sWQf5S0GJE0AKBrFqjpeYcg=="],

-    "@opentelemetry/api-logs": ["@opentelemetry/api-logs@0.214.0", "", { "dependencies": { "@opentelemetry/api": "^1.3.0" } }, "sha512-40lSJeqYO8Uz2Yj7u94/SJWE/wONa7rmMKjI1ZcIjgf3MHNHv1OZUCrCETGuaRF62d5pQD1wKIW+L4lmSMTzZA=="],
+    "@opentelemetry/api-logs": ["@opentelemetry/api-logs@0.219.0", "", { "dependencies": { "@opentelemetry/api": "^1.3.0" } }, "sha512-FFx7YnaYJlIjqWW/AG/yAZ0L/NEY724PipXXXQLdtZPbLwBGbUMTGL1i/esI56TWfTUXxhLfpgrnWJCG8aUJyg=="],

-    "@opentelemetry/context-async-hooks": ["@opentelemetry/context-async-hooks@2.6.1", "", { "peerDependencies": { "@opentelemetry/api": ">=1.0.0 <1.10.0" } }, "sha512-XHzhwRNkBpeP8Fs/qjGrAf9r9PRv67wkJQ/7ZPaBQQ68DYlTBBx5MF9LvPx7mhuXcDessKK2b+DcxqwpgkcivQ=="],
+    "@opentelemetry/context-async-hooks": ["@opentelemetry/context-async-hooks@2.8.0", "", { "peerDependencies": { "@opentelemetry/api": ">=1.0.0 <1.10.0" } }, "sha512-/3FIraneMcng67SUJCxvyInk/oxzwsxyadufk0wwfOBLf5wqtAGX4MoQASwSbndBPeARzBryUM9Azr5kHIdWLw=="],

-    "@opentelemetry/core": ["@opentelemetry/core@2.6.1", "", { "dependencies": { "@opentelemetry/semantic-conventions": "^1.29.0" }, "peerDependencies": { "@opentelemetry/api": ">=1.0.0 <1.10.0" } }, "sha512-8xHSGWpJP9wBxgBpnqGL0R3PbdWQndL1Qp50qrg71+B28zK5OQmUgcDKLJgzyAAV38t4tOyLMGDD60LneR5W8g=="],
+    "@opentelemetry/core": ["@opentelemetry/core@2.8.0", "", { "dependencies": { "@opentelemetry/semantic-conventions": "^1.29.0" }, "peerDependencies": { "@opentelemetry/api": ">=1.0.0 <1.10.0" } }, "sha512-hd1Lfh8p545nNz+jq1Ejfz+Mn1hyLuxYn1YzTfFNrxr8urEWMNQLPf1Th8kjOH+HxwawCrtgBp8JpBUR4ZSgww=="],

-    "@opentelemetry/exporter-trace-otlp-http": ["@opentelemetry/exporter-trace-otlp-http@0.214.0", "", { "dependencies": { "@opentelemetry/core": "2.6.1", "@opentelemetry/otlp-exporter-base": "0.214.0", "@opentelemetry/otlp-transformer": "0.214.0", "@opentelemetry/resources": "2.6.1", "@opentelemetry/sdk-trace-base": "2.6.1" }, "peerDependencies": { "@opentelemetry/api": "^1.3.0" } }, "sha512-kIN8nTBMgV2hXzV/a20BCFilPZdAIMYYJGSgfMMRm/Xa+07y5hRDS2Vm12A/z8Cdu3Sq++ZvJfElokX2rkgGgw=="],
+    "@opentelemetry/exporter-trace-otlp-http": ["@opentelemetry/exporter-trace-otlp-http@0.219.0", "", { "dependencies": { "@opentelemetry/core": "2.8.0", "@opentelemetry/otlp-exporter-base": "0.219.0", "@opentelemetry/otlp-transformer": "0.219.0", "@opentelemetry/resources": "2.8.0", "@opentelemetry/sdk-trace-base": "2.8.0" }, "peerDependencies": { "@opentelemetry/api": "^1.3.0" } }, "sha512-9t6SvBXXBEjOBcIzgozvBbd3jWrv3Gt3ngGhl1fhdZ/zRc7oZDVOFEqbi2zlBpW9BXhgDMKv422J0DL/3iQWfw=="],

    "@opentelemetry/instrumentation": ["@opentelemetry/instrumentation@0.220.0", "", { "dependencies": { "@opentelemetry/api-logs": "0.220.0", "import-in-the-middle": "^3.0.0", "require-in-the-middle": "^8.0.0" }, "peerDependencies": { "@opentelemetry/api": "^1.3.0" } }, "sha512-xQx3E2WxP1mDvKzxLxX+CTCtNLa560YJZ3087qYHerl2YmiKpv7AH+dAy7vmx+eVrZ5BwhfWUAVoKOoxCNHcpw=="],

-    "@opentelemetry/otlp-exporter-base": ["@opentelemetry/otlp-exporter-base@0.214.0", "", { "dependencies": { "@opentelemetry/core": "2.6.1", "@opentelemetry/otlp-transformer": "0.214.0" }, "peerDependencies": { "@opentelemetry/api": "^1.3.0" } }, "sha512-u1Gdv0/E9wP+apqWf7Wv2npXmgJtxsW2XL0TEv9FZloTZRuMBKmu8cYVXwS4Hm3q/f/3FuCnPTgiwYvIqRSpRg=="],
+    "@opentelemetry/otlp-exporter-base": ["@opentelemetry/otlp-exporter-base@0.219.0", "", { "dependencies": { "@opentelemetry/core": "2.8.0", "@opentelemetry/otlp-transformer": "0.219.0" }, "peerDependencies": { "@opentelemetry/api": "^1.3.0" } }, "sha512-zvIxQX/AZUVKDU+hCuYx+7UkiP7GRdnk1ZbFQRYzHvYp47cAWR4j3IhoPhV9KaeXEv2xdGq3IA6PnpzDmLcmSA=="],

-    "@opentelemetry/otlp-transformer": ["@opentelemetry/otlp-transformer@0.214.0", "", { "dependencies": { "@opentelemetry/api-logs": "0.214.0", "@opentelemetry/core": "2.6.1", "@opentelemetry/resources": "2.6.1", "@opentelemetry/sdk-logs": "0.214.0", "@opentelemetry/sdk-metrics": "2.6.1", "@opentelemetry/sdk-trace-base": "2.6.1", "protobufjs": "^7.0.0" }, "peerDependencies": { "@opentelemetry/api": "^1.3.0" } }, "sha512-DSaYcuBRh6uozfsWN3R8HsN0yDhCuWP7tOFdkUOVaWD1KVJg8m4qiLUsg/tNhTLS9HUYUcwNpwL2eroLtsZZ/w=="],
+    "@opentelemetry/otlp-transformer": ["@opentelemetry/otlp-transformer@0.219.0", "", { "dependencies": { "@opentelemetry/api-logs": "0.219.0", "@opentelemetry/core": "2.8.0", "@opentelemetry/resources": "2.8.0", "@opentelemetry/sdk-logs": "0.219.0", "@opentelemetry/sdk-metrics": "2.8.0", "@opentelemetry/sdk-trace-base": "2.8.0" }, "peerDependencies": { "@opentelemetry/api": "^1.3.0" } }, "sha512-aaYKAyXhw9VchKZVGOopD3Gw/kPsyrX2c6IQ0AW32mTjqmZOh5Y6Gf5OYqTNqVktAeBjmFinhyFaCwW6GYK9YQ=="],

-    "@opentelemetry/resources": ["@opentelemetry/resources@2.6.1", "", { "dependencies": { "@opentelemetry/core": "2.6.1", "@opentelemetry/semantic-conventions": "^1.29.0" }, "peerDependencies": { "@opentelemetry/api": ">=1.3.0 <1.10.0" } }, "sha512-lID/vxSuKWXM55XhAKNoYXu9Cutoq5hFdkbTdI/zDKQktXzcWBVhNsOkiZFTMU9UtEWuGRNe0HUgmsFldIdxVA=="],
+    "@opentelemetry/resources": ["@opentelemetry/resources@2.8.0", "", { "dependencies": { "@opentelemetry/core": "2.8.0", "@opentelemetry/semantic-conventions": "^1.29.0" }, "peerDependencies": { "@opentelemetry/api": ">=1.3.0 <1.10.0" } }, "sha512-qmXQ27ilDbUK/vGMqwL8D4/rhn76C+sherM4wTbjlfknR8Nvfc/hCxjRJPhkzZzUsPiNg16SA31NxMabwttRjg=="],

-    "@opentelemetry/sdk-logs": ["@opentelemetry/sdk-logs@0.214.0", "", { "dependencies": { "@opentelemetry/api-logs": "0.214.0", "@opentelemetry/core": "2.6.1", "@opentelemetry/resources": "2.6.1", "@opentelemetry/semantic-conventions": "^1.29.0" }, "peerDependencies": { "@opentelemetry/api": ">=1.4.0 <1.10.0" } }, "sha512-zf6acnScjhsaBUU22zXZ/sLWim1dfhUAbGXdMmHmNG3LfBnQ3DKsOCITb2IZwoUsNNMTogqFKBnlIPPftUgGwA=="],
+    "@opentelemetry/sdk-logs": ["@opentelemetry/sdk-logs@0.219.0", "", { "dependencies": { "@opentelemetry/api-logs": "0.219.0", "@opentelemetry/core": "2.8.0", "@opentelemetry/resources": "2.8.0", "@opentelemetry/semantic-conventions": "^1.29.0" }, "peerDependencies": { "@opentelemetry/api": ">=1.4.0 <1.10.0" } }, "sha512-s6lTKRakaPClvKoWHRChxnXjDMkM/TQ30ff78jN6EBGf7MI7VzANE5PU3f4z9qDUudWjvZjOLHG0rBnBKYvoXA=="],

-    "@opentelemetry/sdk-metrics": ["@opentelemetry/sdk-metrics@2.6.1", "", { "dependencies": { "@opentelemetry/core": "2.6.1", "@opentelemetry/resources": "2.6.1" }, "peerDependencies": { "@opentelemetry/api": ">=1.9.0 <1.10.0" } }, "sha512-9t9hJHX15meBy2NmTJxL+NJfXmnausR2xUDvE19XQce0Qi/GBtDGamU8nS1RMbdgDmhgpm3VaOu2+fiS/SfTpQ=="],
+    "@opentelemetry/sdk-metrics": ["@opentelemetry/sdk-metrics@2.8.0", "", { "dependencies": { "@opentelemetry/core": "2.8.0", "@opentelemetry/resources": "2.8.0" }, "peerDependencies": { "@opentelemetry/api": ">=1.9.0 <1.10.0" } }, "sha512-UDBGaj6W0Rgy5rTTaoxs8gVGF/aGkAKyjurJv7se6wjRxJu7FoquTLT/vt54DZfo4crbprYfhX/SOK9+BPw1qg=="],

    "@opentelemetry/sdk-trace": ["@opentelemetry/sdk-trace@2.11.0", "", { "dependencies": { "@opentelemetry/core": "2.11.0", "@opentelemetry/resources": "2.11.0", "@opentelemetry/semantic-conventions": "^1.29.0" }, "peerDependencies": { "@opentelemetry/api": ">=1.3.0 <1.10.0" } }, "sha512-fFnTqGm8/G73GQVnxYi7LXa1ZVYEUvgL6XI1LpvV0bPC7WQ/ZGgKxCSl8FnlZBKto9JHHEFTO6s6CUpvvtwFrA=="],

-    "@opentelemetry/sdk-trace-base": ["@opentelemetry/sdk-trace-base@2.6.1", "", { "dependencies": { "@opentelemetry/core": "2.6.1", "@opentelemetry/resources": "2.6.1", "@opentelemetry/semantic-conventions": "^1.29.0" }, "peerDependencies": { "@opentelemetry/api": ">=1.3.0 <1.10.0" } }, "sha512-r86ut4T1e8vNwB35CqCcKd45yzqH6/6Wzvpk2/cZB8PsPLlZFTvrh8yfOS3CYZYcUmAx4hHTZJ8AO8Dj8nrdhw=="],
+    "@opentelemetry/sdk-trace-base": ["@opentelemetry/sdk-trace-base@2.8.0", "", { "dependencies": { "@opentelemetry/core": "2.8.0", "@opentelemetry/resources": "2.8.0", "@opentelemetry/semantic-conventions": "^1.29.0" }, "peerDependencies": { "@opentelemetry/api": ">=1.3.0 <1.10.0" } }, "sha512-mhU4jp+vW0mGbFRd+GeXHvmfA4aDqWjBjLC3pE5XMpLs0IE2ryYb019Ts2AQrOq67gaTF25D91+fgvEHDZEnuQ=="],

-    "@opentelemetry/sdk-trace-node": ["@opentelemetry/sdk-trace-node@2.6.1", "", { "dependencies": { "@opentelemetry/context-async-hooks": "2.6.1", "@opentelemetry/core": "2.6.1", "@opentelemetry/sdk-trace-base": "2.6.1" }, "peerDependencies": { "@opentelemetry/api": ">=1.0.0 <1.10.0" } }, "sha512-Hh2i4FwHWRFhnO2Q/p6svMxy8MPsNCG0uuzUY3glqm0rwM0nQvbTO1dXSp9OqQoTKXcQzaz9q1f65fsurmOhNw=="],
+    "@opentelemetry/sdk-trace-node": ["@opentelemetry/sdk-trace-node@2.8.0", "", { "dependencies": { "@opentelemetry/context-async-hooks": "2.8.0", "@opentelemetry/core": "2.8.0", "@opentelemetry/sdk-trace-base": "2.8.0" }, "peerDependencies": { "@opentelemetry/api": ">=1.0.0 <1.10.0" } }, "sha512-nZt9OGufioAc3AfoLTqA9bsAeaMJAictYDdI2VcNQ+PmT+3rfKjAZDZvgPfd8VPX0O5Bw1hdQF6kDK8VSpZiWg=="],

    "@opentelemetry/semantic-conventions": ["@opentelemetry/semantic-conventions@1.43.0", "", {}, "sha512-eSYWTm620tTk45EKSedaUL8MFYI8hW164hIXsgIHyxu3VobUB3fFCu5t0hQby6OoWRPsG1KkKUG2M5UadiLiVg=="],

@@ -2554,24 +2558,6 @@

    "@protobuf-ts/runtime-rpc": ["@protobuf-ts/runtime-rpc@2.11.1", "", { "dependencies": { "@protobuf-ts/runtime": "^2.11.1" } }, "sha512-4CqqUmNA+/uMz00+d3CYKgElXO9VrEbucjnBFEjqI4GuDrEQ32MaI3q+9qPBvIGOlL4PmHXrzM32vBPWRhQKWQ=="],

-    "@protobufjs/aspromise": ["@protobufjs/aspromise@1.1.2", "", {}, "sha512-j+gKExEuLmKwvz3OgROXtrJ2UG2x8Ch2YZUxahh+s1F2HZ+wAceUNLkvy6zKCPVRkU++ZWQrdxsUeQXmcg4uoQ=="],
-
-    "@protobufjs/base64": ["@protobufjs/base64@1.1.2", "", {}, "sha512-AZkcAA5vnN/v4PDqKyMR5lx7hZttPDgClv83E//FMNhR2TMcLUhfRUBHCmSl0oi9zMgDDqRUJkSxO3wm85+XLg=="],
-
-    "@protobufjs/codegen": ["@protobufjs/codegen@2.0.5", "", {}, "sha512-zgXFLzW3Ap33e6d0Wlj4MGIm6Ce8O89n/apUaGNB/jx+hw+ruWEp7EwGUshdLKVRCxZW12fp9r40E1mQrf/34g=="],
-
-    "@protobufjs/eventemitter": ["@protobufjs/eventemitter@1.1.1", "", {}, "sha512-vW1GmwMZNnL+gMRaovlh9yZX74kc+TTU3FObkkurpMaRtBfLP3ldjS9KQWlwZgraRE0+dheEEoAxdzcJQ8eXZg=="],
-
-    "@protobufjs/fetch": ["@protobufjs/fetch@1.1.1", "", { "dependencies": { "@protobufjs/aspromise": "^1.1.1" } }, "sha512-GpptLrs57adMSuHi3VNj0mAF8dwh36LMaYF6XyJ6JMWlVsc+t42tm1HSEDmOs3A8fC9yyeisgLhsTVQokOZ0zw=="],
-
-    "@protobufjs/float": ["@protobufjs/float@1.0.2", "", {}, "sha512-Ddb+kVXlXst9d+R9PfTIxh1EdNkgoRe5tOX6t01f1lYWOvJnSPDBlG241QLzcyPdoNTsblLUdujGSE4RzrTZGQ=="],
-
-    "@protobufjs/path": ["@protobufjs/path@1.1.2", "", {}, "sha512-6JOcJ5Tm08dOHAbdR3GrvP+yUUfkjG5ePsHYczMFLq3ZmMkAD98cDgcT2iA1lJ9NVwFd4tH/iSSoe44YWkltEA=="],
-
-    "@protobufjs/pool": ["@protobufjs/pool@1.1.0", "", {}, "sha512-0kELaGSIDBKvcgS4zkjz1PeddatrjYcmMWOlAuAPwAeccUrPHdUqo/J6LiymHHEiJT5NrF1UVwxY14f+fy4WQw=="],
-
-    "@protobufjs/utf8": ["@protobufjs/utf8@1.1.2", "", {}, "sha512-b1UQwcEZ4yCnMCD8DAL1VlbvBJE9/IX4FTIp7BG1xYpf29SLazLSrqUkj4w7Y5y7cCVP6E5tcqqcI0xemPkHug=="],
-
    "@puppeteer/browsers": ["@puppeteer/browsers@3.2.1", "", { "dependencies": { "modern-tar": "^0.8.0", "yargs": "^18.0.0" }, "peerDependencies": { "proxy-agent": ">=8.0.1", "yauzl": "^2.10.0 || ^3.4.0" }, "optionalPeers": ["proxy-agent", "yauzl"], "bin": { "browsers": "lib/main-cli.js" } }, "sha512-KDz+3qDRdBAlRlMjmKyj6dEs33YHTk/xRHEENSXq6TNnhgoU15ruSHtEBeVF6OZ9tBDY55Se4P0nFMNsipzU9A=="],

    "@radix-ui/colors": ["@radix-ui/colors@1.0.1", "", {}, "sha512-xySw8f0ZVsAEP+e7iLl3EvcBXX7gsIlC1Zso/sPBW9gIWerBTgz6axrjU+MZ39wD+WFi5h5zdWpsg3+hwt2Qsg=="],
@@ -5048,8 +5034,6 @@

    "proto-list": ["proto-list@1.2.4", "", {}, "sha512-vtK/94akxsTMhe0/cbfpR+syPuszcuwhqVjJq26CuNDgFGj682oRBXOP5MJpv2r7JtE8MsiepGIqvvOTBwn2vA=="],

-    "protobufjs": ["protobufjs@7.6.5", "", { "dependencies": { "@protobufjs/aspromise": "^1.1.2", "@protobufjs/base64": "^1.1.2", "@protobufjs/codegen": "^2.0.5", "@protobufjs/eventemitter": "^1.1.1", "@protobufjs/fetch": "^1.1.1", "@protobufjs/float": "^1.0.2", "@protobufjs/path": "^1.1.2", "@protobufjs/pool": "^1.1.0", "@protobufjs/utf8": "^1.1.1", "@types/node": ">=13.7.0", "long": "^5.3.2" } }, "sha512-/FPD0nUc9jH6rfFjji9IBqOz4pcSE3CsT1m7Ep6Mdb0LxSUMj8hgl6GomOvZzpNpAqqGaXA0P3VSrZLFzIhQrw=="],
-
    "proxy-from-env": ["proxy-from-env@1.1.0", "", {}, "sha512-D+zkORCbA9f1tdWRK0RaCR3GPv50cMxcrz4X8k5LTSUD1Dkw47mKJEZQNunItRTkWwgtaUSo1RVFRIG9ZXiFYg=="],

    "pump": ["pump@3.0.4", "", { "dependencies": { "end-of-stream": "^1.1.0", "once": "^1.3.1" } }, "sha512-VS7sjc6KR7e1ukRFhQSY5LM2uBWAUPiOPa/A3mkKmiMwSmRFUITt0xuj+/lesgnCv+dPIEYlkzrcyXgquIHMcA=="],
@@ -6246,6 +6230,8 @@

    "@opencode/www/wrangler": ["wrangler@4.110.0", "", { "dependencies": { "@cloudflare/kv-asset-handler": "0.5.0", "@cloudflare/unenv-preset": "2.16.1", "blake3-wasm": "2.1.5", "esbuild": "0.28.1", "miniflare": "4.20260708.1", "path-to-regexp": "6.3.0", "unenv": "2.0.0-rc.24", "workerd": "1.20260708.1" }, "optionalDependencies": { "fsevents": "2.3.3" }, "peerDependencies": { "@cloudflare/workers-types": "^5.20260708.1" }, "optionalPeers": ["@cloudflare/workers-types"], "bin": { "wrangler": "bin/wrangler.js", "wrangler2": "bin/wrangler.js", "cf-wrangler": "bin/cf-wrangler.js" } }, "sha512-xZeXKYi7hxQRF5anL+v77RkufJNpF9f3Eqeyqq2QBsETpLZgh0Agj0jJ6JPtkbgn6ukZdh8OK5egsGPWIditgg=="],

+    "@opentelemetry/api-logs/@opentelemetry/api": ["@opentelemetry/api@1.9.1", "", {}, "sha512-gLyJlPHPZYdAk1JENA9LeHejZe1Ti77/pTeFm/nMXmQH/HFZlcS/O2XJB+L8fkbrNSqhdtlvjBVjxwUYanNH5Q=="],
+
    "@opentelemetry/instrumentation/@opentelemetry/api-logs": ["@opentelemetry/api-logs@0.220.0", "", { "dependencies": { "@opentelemetry/api": "^1.3.0" } }, "sha512-CmVa4ImJ+ynfrPMNaAXHET6Bhb44SwzmfyVJFq9ni2jgXJR/l7C6gfVFddNmHP+ZOkP9cf4f9DBe68qVLTHc9w=="],

    "@opentelemetry/sdk-trace/@opentelemetry/core": ["@opentelemetry/core@2.11.0", "", { "dependencies": { "@opentelemetry/semantic-conventions": "^1.29.0" }, "peerDependencies": { "@opentelemetry/api": ">=1.0.0 <1.10.0" } }, "sha512-7YP44XH0tV6+Mb54x2YGf84i7yi+31MBZlE8JwvozkxyTvXbSp10X7cI7YE49ChJ3shMJoBmCJF3+1QFBJctGA=="],
@@ -1,8 +1,8 @@
 {
  "nodeModules": {
-    "x86_64-linux": "sha256-DVimx64BQEkFaYZkW56E5eEfkGxuW5j/9C0wCImBFl4=",
-    "aarch64-linux": "sha256-MhMl/ZkvcPXE31OvysutI6nKIJ6XtUa3c71+B4hnaQc=",
-    "aarch64-darwin": "sha256-+wMPYLBnw2zTZj0if8EPwWC9O3tVqSgnhYNfQq8UZ8Q=",
-    "x86_64-darwin": "sha256-l6LxVSXDbqVoXMVcGZAzRER0bxP6wDU4dfRDolx0JGM="
+    "x86_64-linux": "sha256-LQ1GAz1qF4R5P4j/kkgUygsQvxm7KStVlMf24nmyq44=",
+    "aarch64-linux": "sha256-PsNR3VaClA1O1vSE0z/Hb3XRT1pb6PvNRXuK+giIx6s=",
+    "aarch64-darwin": "sha256-7VCS+GT7tDYAMgqNCnLh1zvHAZkLTnJJ4vtfM88ruT8=",
+    "x86_64-darwin": "sha256-HyxXjiFVd8vcKvSb/BuHAY5Jj3MGgdXEvuhDuglp4ss="
  }
 }
@@ -2,7 +2,7 @@
  "$schema": "https://json.schemastore.org/package.json",
  "name": "opencode",
  "description": "AI-powered development tool",
-  "version": "2.0.14",
+  "version": "2.0.15",
  "private": true,
  "type": "module",
  "packageManager": "bun@1.4.2",
@@ -96,7 +96,13 @@ When a provider supports multiple physical transports, selection remains executi

 ### Media Routes

-Media does not fit the SSE-frames-to-event-state-machine LLM route. `MediaRoute.make(...)` (`src/route/media.ts`) composes a `MediaProtocol` kind with `Endpoint` and `Auth` and owns the transport plumbing: `http` option merging, URL/query rendering, auth headers, JSON vs multipart encoding, and handing the response back to the protocol. `MediaProtocol.inline` (`src/route/media-protocol.ts`) is `body.from(request)` plus `response.decode(response, context)`; use `MediaProtocol.decodeJson` / `text` / `bytes` so decode failures retain the raw body and HTTP context. `Generation` (`src/generation.ts`) is the provider-neutral handle for a queued generation over a `GenerationRoute` (`status`, `result`, `cancel`, `pollHint`); the first video route implements it. Image protocol files follow the same section order as LLM protocols and declare unsupported common fields once through `MediaInput.rejectUnsupported`.
+Media does not fit the SSE-frames-to-event-state-machine LLM route. `MediaRoute.make(...)` (`src/route/media.ts`) composes a `MediaProtocol` kind with `Endpoint` and `Auth` and owns the transport plumbing: `http` option merging, URL/query rendering, auth headers, JSON vs multipart encoding, and handing the response back to the protocol. `MediaProtocol.inline` (`src/route/media-protocol.ts`) is `body.from(request)` plus `response.decode(response, context)`; use `MediaProtocol.decodeJson` / `text` / `bytes` so decode failures retain the raw body and HTTP context. `Generation` (`src/generation.ts`) is the provider-neutral handle for a queued generation over a `GenerationRoute` (`status`, `result`, `cancel`, `pollHint`). Image protocol files follow the same section order as LLM protocols and declare unsupported common fields once through `MediaInput.rejectUnsupported`.
+
+`MediaProtocol.queued` is the submit-then-poll kind every video route uses: `start` (body + decode into `{ token, snapshot }`), `status`, `result`, and optional `cancel`, each addressed by a route-owned `token` whose `Schema.Codec` makes it serializable. `MediaRoute.inline` and `MediaRoute.queued` compose the two kinds with `Endpoint` and `Auth`; the queued route decodes the token once at the boundary (`start` output or `resume` input) and closes over it in a token-free `GenerationRoute` (`status`/`result`/`cancel` are plain Effects), so `Generation` never sees the token's shape and only carries the encoded JSON for persistence. Polls reuse the route's auth and deployment headers plus the request's `http` overlay after `start`, and resolve relative paths against the route base URL (provider-issued absolute URLs such as fal's `status_url` pass through). `result` is always its own GET even when the provider returns output inside the status document, so `Generation.await` behaves the same after `start` and after `resume`. `PollContext.auth` carries only what `Auth` added so protocols can hand download credentials to output assets as transient `Media.Asset.headers` (Veo) — never part of `source` or JSON. Status strings map through a per-protocol `STATUS` table via `MediaProtocol.status`; terminal generations without output fail through `output.ended` / `output.contentPolicy` with the provider document on `reason.body`. `Generation.AwaitOptions` (`{ poll?: Poll }`) is the one options type for `await`, `events`, `Video.generate`, and `Video.stream`.
+
+`MediaProtocol.stream` is the incremental kind every speech route uses, with the same discipline as LLM protocols. `MediaRoute.stream` submits the caller's request as `MediaProtocol.Addressed<Request>` (`{ ...request, mode }`, `mode: "generate" | "stream"`), so one provider stays one protocol: `body.from`, the endpoint path, and `frames` read `request.mode` to pick the body, path, and framing. `frames(bytes, context)` returns frames — `Framing.sse`, `Framing.lines`, `Framing.document` (a single-document response shaped like a streamed record), or the raw `bytes` for chunked audio. `initial()` is fresh per-response parser state; `step` folds each frame into it and emits modality events; `finish(state, context)` runs once after the last frame with the request, body, and observed `http` (header-only usage lives there) and emits exactly one terminal event or fails with `MediaProtocol.incomplete`. Keep parser state to real accumulators and derive anything the request or body determines in `finish`. `generate` runs the same stream and folds it with the modality's `collect`. Request-derived URL parameters go on the body's `query` (array values repeat the parameter), applied before route and caller `http.query`. Decode frames with `MediaProtocol.decodeFrame` and raise stream-time failures with `MediaProtocol.frameError` (the frame stays on `reason.body`); protocols never thread HTTP context, because the route fills `reason.http` on stream errors that lack it. Speech protocols share `protocols/utils/speech-stream.ts` for deltas, timestamps, voice ids, PCM and container descriptions, and the terminal asset.
+
+Transcription uses all three kinds (OpenAI and Gemini stream, Deepgram is inline, AssemblyAI is queued): every route carries its `kind` and `TranscriptionClient` dispatches on it. Bodies are `json`, `multipart`, or `binary` (a raw upload), and a queued protocol that must upload media before submitting implements `start.prepare` (`MediaProtocol.Prepare`; AssemblyAI `/v2/upload`).

 ### URL Construction

@@ -602,6 +602,233 @@ const program = Effect.gen(function* () {

 The hosted result is represented as a provider-executed tool call and tool result, and the generated image is also emitted as a first-class `media` `LLMEvent` (`response.message` then carries a `media` part). Gemini image-capable models emit the same `media` event for inline image output. Retaining `response.message` preserves the generated image for continuation on both routes.

+## Video generation
+
+Video mirrors `Image` with one difference: every provider is asynchronous, so the route is a submit-then-poll
+`Generation`. Models come from `.video(...)` selectors on the `Google` (Veo), `XAI`, `Fal`, and `Runway` facades.
+Common fields (`frames`, `references`, `video`, `durationSeconds`, `aspectRatio`, `resolution`, `audio`, `n`, `seed`,
+`negativePrompt`) lower natively or fail with a typed `AIError` before any network call; provider-native controls live
+under `providerOptions`, inferred from the selected model.
+
+```ts
+import { Video, VideoClient } from "@opencode/ai"
+import { Google } from "@opencode/ai/providers"
+
+const google = Google.configure({ apiKey: process.env.GOOGLE_GENERATIVE_AI_API_KEY })
+
+// Simple: submit and wait.
+const program = Effect.gen(function* () {
+  const response = yield* Video.generate(
+    {
+      model: google.video("veo-3.1-generate-preview"),
+      prompt: "Panning wide shot of a calico kitten sleeping in the sunshine",
+      aspectRatio: "16:9",
+      resolution: "1080p",
+      durationSeconds: 8,
+      providerOptions: { personGeneration: "allow_adult" },
+    },
+    { poll: { interval: "10 seconds", timeout: "10 minutes" } },
+  )
+  // Veo serves files for two days behind the API key. The asset knows the deadline (`expiresAt`) and carries the
+  // download credentials only on the live instance (`asset.headers`), never in `source` or JSON: materialize
+  // before persisting, or the persisted URL cannot be fetched again.
+  return yield* response.video.materialize()
+})
+
+// Explicit control: keep the handle, persist the token, resume elsewhere.
+const controlled = Effect.gen(function* () {
+  const generation = yield* Video.start({ model: google.video("veo-3.1-generate-preview"), prompt })
+  generation.id // provider operation / task / request id
+  generation.status // "queued" | "running" | "completed" | "failed" | "cancelled" | "expired"
+  generation.token // route-owned JSON: `{ operation }`, `{ requestID }`, `{ taskID }`, or fal's follow-up URLs
+  const saved = JSON.stringify(generation.token)
+
+  const resumed = yield* Video.resume(google.video("veo-3.1-generate-preview"), JSON.parse(saved))
+  return yield* resumed.await({ poll: { interval: "10 seconds" } })
+})
+
+// Progress as a stream: generation-queued | generation-progress | video | finish.
+const events = Video.stream({ model: Runway.configure({ apiKey }).video("gen4.5"), prompt }, { poll })
+```
+
+`VideoClient.layer` needs `RequestExecutor.Service`, and status polls, result fetches, cancels, and asset downloads
+all run through the same executor with the route's auth. `Generation.await` and `Generation.events` fail with a
+`Timeout` reason when `poll.timeout` (default 10 minutes) elapses. Failed,
+cancelled, and expired generations fail typed with the provider's terminal document on `reason.body`; moderation
+outcomes (Veo `raiMediaFilteredReasons`, xAI `respect_moderation`, Runway `SAFETY.*` codes) surface as `notices` when
+a video is still returned and as a `ContentPolicy` reason when nothing is.
+
+Provider notes:
+
+- **Google Veo** takes inline bytes only (materialize `url` assets first); `frames.last` requires `frames.first`;
+  audio is always on, so `audio: false` fails typed; one video per request. Output URLs need the API key to
+  download, which the returned asset holds transiently (see above).
+- **xAI** sends a `video` input to `/videos/edits`, or `/videos/extensions` with `providerOptions.mode: "extend"`.
+  `seed` and `negativePrompt` are not supported.
+- **fal** endpoints are model-specific: `durationSeconds`, `references`, and `frames.last` fail typed and belong in
+  `providerOptions` under the model's own names (`duration: "8s"`, `end_image_url`, …). Auth is
+  `Authorization: Key <FAL_KEY>`.
+- **Runway** expects pixel ratios in `aspectRatio` for most models (`"1280:720"`), pins `X-Runway-Version`, reports
+  `usage: { type: "credits" }`, and its output URLs expire after 24–48 hours.
+
+The promise client exposes the same surface: `ai.video.start(...)` resolves to a handle with `await`, `refresh`,
+`cancel`, and `token`; `ai.video.generate`, `ai.video.resume(model, token)`, and `ai.video.stream` mirror the Effect
+API.
+
+```ts
+import { ai } from "@opencode/ai/promise"
+
+const generation = await ai.video.start({ model, prompt })
+const video = await generation.await({ poll: { interval: 10_000 }, signal })
+```
+
+## Speech generation
+
+Speech (text-to-speech) is one request whose response is parsed incrementally, so every route supports both
+`Speech.generate` (the whole file) and `Speech.stream` (audio chunks as they arrive). Models come from `.speech(...)`
+selectors on the `OpenAI`, `Google` (Gemini TTS), `ElevenLabs`, `Cartesia`, and `Deepgram` facades. Common fields
+(`voice`, `format`, `speed`, `language`, `instructions`, `timestamps`) lower natively or fail with a typed `AIError`
+before any network call; provider-native controls live under `providerOptions`, inferred from the selected model.
+
+```ts
+import { Media, Speech, SpeechClient, SpeechEvent } from "@opencode/ai"
+import { ElevenLabs, OpenAI } from "@opencode/ai/providers"
+
+const openai = OpenAI.configure({ apiKey: process.env.OPENAI_API_KEY })
+
+// The whole file, written to disk.
+const program = Effect.gen(function* () {
+  const response = yield* Speech.generate({
+    model: openai.speech("gpt-4o-mini-tts"),
+    text: "Hello from OpenCode.",
+    voice: "coral",
+    format: "mp3",
+    instructions: "Warm and unhurried.",
+  })
+  response.audio // Media.Asset with bytes; headerless PCM carries info.encoding / sampleRate / channels
+  response.usage // undefined: OpenAI reports tokens only on SSE streams (Gemini: tokens; ElevenLabs: credits; Deepgram: characters)
+  yield* Media.write(response.audio, "hello.mp3")
+})
+
+// Chunks as they arrive: audio-delta* (interleaved with timestamps) then one finish carrying the assembled asset.
+const events = Speech.stream({
+  model: ElevenLabs.configure({ apiKey }).speech("eleven_flash_v2_5"),
+  text: "Hello from OpenCode.",
+  voice: "JBFqnCBsd6RMkjVDRZzb",
+  format: "pcm",
+  timestamps: true,
+}).pipe(
+  Stream.tap((event) => {
+    if (SpeechEvent.is.audioDelta(event)) return play(event.chunk)
+    if (SpeechEvent.is.timestamps(event)) return highlight(event.items) // { text, startSeconds, endSeconds }[]
+    return Effect.void
+  }),
+)
+```
+
+`voice` is the provider's own identifier — a name on OpenAI and Gemini (`"coral"`, `"Kore"`), a voice id on
+ElevenLabs and Cartesia. `{ id }` selects an OpenAI custom voice (`{ id: "voice_1234" }`) and means the same as the
+plain string elsewhere. There is no cross-provider voice catalog. `format` is the container-level word (`mp3`, `wav`,
+`pcm`, `opus`, `aac`, `flac`); sample rates and bitrates live under `providerOptions`, and a value the route cannot
+produce fails as `UnsupportedOperation`. Streams buffer every chunk so `finish` can carry the whole clip.
+`SpeechClient.layer` needs `RequestExecutor.Service`.
+
+Provider notes:
+
+- **OpenAI** streams over SSE (`stream_format: "sse"`), which is also the only place it reports token usage; `tts-1`
+  and `tts-1-hd` do not support SSE and stream the raw audio body instead. `pcm` is 24 kHz 16-bit mono. `language`
+  and `timestamps` are not supported.
+- **Gemini TTS** returns raw 16-bit PCM only (`audio/L16;codec=pcm;rate=24000`), so any `format` other than `pcm`
+  fails typed; wrap the samples yourself. Style is directed in the text, so `instructions` and `speed` fail typed.
+  Only `gemini-3.1-flash-tts-preview` and later support streaming. Two-speaker audio goes through
+  `providerOptions.speechConfig.multiSpeakerVoiceConfig`.
+- **ElevenLabs** requires `voice` (the path voice id) and authenticates with `xi-api-key`. `format` maps to the
+  `output_format` query parameter (`mp3_44100_128`, `pcm_24000`, `wav_24000`, `opus_48000_64`);
+  `providerOptions.outputFormat` sets the exact string. WAV is only available from `generate`. `timestamps: true`
+  selects the `with-timestamps` endpoints and yields character-level alignment. `instructions` is not supported.
+- **Cartesia** requires `voice` and pins `Cartesia-Version`. `generate` defaults to MP3 from `/tts/bytes`; streams
+  and `timestamps: true` (word-level) use `/tts/sse`, which only serves raw PCM. `providerOptions.sampleRate`,
+  `bitRate`, and `encoding` complete `output_format`. No usage is reported.
+- **Deepgram** Aura's voice is the model id (`aura-2-thalia-en`), so `voice` and `language` fail typed. `format`
+  and `providerOptions` lower to query parameters (`encoding`, `container`, `sample_rate`, `bit_rate`); `pcm` is
+  `linear16` without a container. Auth is `Authorization: Token <DEEPGRAM_API_KEY>`.
+
+The promise client mirrors the Effect API; `ai.speech.stream` is an `AsyncIterable`.
+
+```ts
+import { ai } from "@opencode/ai/promise"
+
+const response = await ai.speech.generate({ model, text: "Hello from OpenCode.", voice: "coral" })
+await Bun.write("hello.mp3", await ai.run(response.audio.bytes()))
+
+for await (const event of ai.speech.stream({ model, text: "Hello from OpenCode.", voice: "coral" })) {
+  if (event.type === "audio-delta") player.write(event.chunk)
+}
+```
+
+## Transcription
+
+Transcription (speech-to-text) is the one modality whose providers use every route kind: OpenAI and Gemini stream,
+Deepgram answers inline, and AssemblyAI is queued. `Transcription.generate` and `Transcription.stream` work on all of
+them; `Transcription.start` / `resume` return a `Generation` on queued routes and fail with `UnsupportedOperation`
+elsewhere. Models come from `.transcription(...)` selectors on the `OpenAI`, `Google`, `Deepgram`, and `AssemblyAI`
+facades. Common fields (`language`, `prompt`, `timestamps: "none" | "segment" | "word"`, `diarize`, `speakers`) lower
+natively or fail with a typed `AIError` before any network call; a route may return more than asked.
+
+```ts
+import { Media, Transcription, TranscriptionEvent } from "@opencode/ai"
+import { AssemblyAI, Deepgram, OpenAI } from "@opencode/ai/providers"
+
+const openai = OpenAI.configure({ apiKey: process.env.OPENAI_API_KEY })
+
+const program = Effect.gen(function* () {
+  const audio = yield* Media.file("./call.mp3")
+
+  // Speaker-labelled segments; labels are provider-native strings ("A", "0", "spk:0").
+  const response = yield* Transcription.generate({
+    model: Deepgram.configure({ apiKey }).transcription("nova-3"),
+    audio,
+    diarize: true,
+    timestamps: "word",
+  })
+  response.text // "Hello from OpenCode."
+  response.segments // [{ text, startSeconds, endSeconds, speaker: "0" }]
+  response.words // [{ text, startSeconds, endSeconds, speaker, confidence }]
+  response.language // the provider's own value, lowercased ("en", "english", "en_us")
+
+  // Text deltas as the model transcribes, then one finish carrying the whole transcript.
+  yield* Transcription.stream({ model: openai.transcription("gpt-4o-mini-transcribe"), audio }).pipe(
+    Stream.tap((event) => (TranscriptionEvent.is.textDelta(event) ? Console.log(event.delta) : Effect.void)),
+    Stream.runDrain,
+  )
+
+  // Queued: persist the token, resume from another process, and await.
+  const model = AssemblyAI.configure({ apiKey }).transcription("universal-3-5-pro")
+  const generation = yield* Transcription.start({ model, audio })
+  const resumed = yield* Transcription.resume(model, JSON.parse(JSON.stringify(generation.token)))
+  const transcript = yield* resumed.await({ poll: { interval: "3 seconds" } })
+})
+```
+
+Inline routes emit only `finish` from `stream` (no faked deltas); queued routes emit `generation-queued` /
+`generation-progress` before it. `TranscriptionClient.layer` needs `RequestExecutor.Service`.
+
+Provider notes:
+
+- **OpenAI** takes inline audio only; `diarize` needs `gpt-4o-transcribe-diarize`, timestamps need `whisper-1`, and `whisper-1` does not stream.
+- **Gemini** needs a transcribe model (`gemini-3.5-transcribe`); `prompt` and `speakers` fail typed.
+- **Deepgram** detects the language unless `language` is set; vocabulary goes in `providerOptions.keyterm`.
+- **AssemblyAI** uploads inline audio before submitting and is the only route that accepts `speakers`.
+
+The promise client mirrors the Effect API:
+
+```ts
+const text = (await ai.transcription.generate({ model, audio })).text
+for await (const event of ai.transcription.stream({ model, audio })) if (event.type === "text-delta") write(event.delta)
+const generation = await ai.transcription.start({ model: assemblyai, audio })
+const transcript = await generation.await({ poll: { interval: 3_000 } })
+```
+
 ## Public API

 - **`LLM.request({...})`** — build a provider-neutral `LLMRequest`. Accepts ergonomic inputs (`system: string`, `prompt: string`) that normalize into the canonical Schema classes.
@@ -613,7 +840,9 @@ The hosted result is represented as a provider-executed tool call and tool resul
 - **`ImageClient`** — Effect service and layer for image execution, parallel to `LLMClient`.
 - **`Media`** — the shared asset type (`Media.Asset`, `Media.Source`) and constructors used by messages, tool results, and media requests.
 - **`Generation`** — provider-neutral handle for an in-flight media generation (`await`, `refresh`, `cancel`, `events`) used by queued media routes.
- **`@opencode/ai/promise`** — `AI.make({ layer? })` and a default `ai` client exposing `llm` and `image` as Promise / `AsyncIterable` APIs.
+- **`Speech.request` / `Speech.generate` / `Speech.stream`** — text-to-speech through a provider-neutral request; `SpeechClient` is its Effect service and layer.
+- **`Transcription.request` / `generate` / `stream` / `start` / `resume`** — speech-to-text over inline, streaming, and queued routes; `TranscriptionClient` is its Effect service and layer.
+- **`@opencode/ai/promise`** — `AI.make({ layer? })` and a default `ai` client exposing `llm`, `image`, `video`, `speech`, and `transcription` as Promise / `AsyncIterable` APIs.

 ## Testing

@@ -1,6 +1,6 @@
 # Media generation in `@opencode/ai` — public API direction

-Status: proposal. Branch `media-support`.
+Status: phases 1–3 implemented (Speech and Transcription); phases 4–5 proposal.

 ## Goal

@@ -66,7 +66,7 @@ import { Media } from "@opencode/ai"
 Media.Source =
  | { type: "bytes";  data: Uint8Array; mediaType: string }
  | { type: "base64"; data: string;     mediaType: string }
-  | { type: "url";    url: string; mediaType?: string; expiresAt?: number; headers?: Record<string, string> }
+  | { type: "url";    url: string; mediaType?: string; expiresAt?: number }
  | { type: "ref";    provider: ProviderID; id: string; mediaType?: string }   // file_id, gs://, runway://, prior generation

 class Media.Asset {
@@ -76,6 +76,7 @@ class Media.Asset {
  readonly info?: { width?; height?; durationSeconds?; sampleRate?; channels?; encoding?; format? }
  readonly expiresAt?: number
  readonly providerMetadata?: ProviderMetadata
+  readonly headers?: Record<string, string>     // transient download credentials (Veo); never in source/JSON

  bytes(): Effect<Uint8Array, AIError, RequestExecutor.Service>   // downloads/decodes lazily, cached
  base64(): Effect<string, AIError, RequestExecutor.Service>
@@ -129,98 +130,210 @@ Editing is not a separate function; `images`/`mask` on the request select the ed

 #### Video

+Shipped in phase 2 (`src/video.ts`, `src/video-client.ts`, protocols `google-video`, `xai-video`, `fal-video`, `runway-video`).
+
 ```ts
 const request = Video.request({
  model: google.video("veo-3.1-generate-preview"),
  prompt: "Panning wide shot of a calico kitten sleeping in the sunshine",
  frames: { first: Media.file("./start.png"), last: Media.file("./end.png") },
-  references: [Media.url("https://…/style.png")],
-  video: Media.ref("openai", "video_123"),      // edit / extend / remix source
+  references: [Media.file("./style.png")],
+  video: Media.bytes(previous, "video/mp4"),    // edit / extend source
  durationSeconds: 8,
  aspectRatio: "16:9",
  resolution: "1080p",
  audio: true,
  n: 1,
-  providerOptions: { personGeneration: "dont_allow", negativePrompt: "text, watermark" },
+  seed: 7,
+  negativePrompt: "text, watermark",            // common, not provider-native
+  providerOptions: { personGeneration: "allow_adult" },
 })

 // Simple: wait for it.
 const response = yield* Video.generate(request, { poll: { interval: "10 seconds", timeout: "10 minutes" } })
-response.video                                     // Media.Asset (url with expiresAt, or bytes when the route downloads)
+response.video                                     // Media.Asset: url with expiresAt (+ transient `headers` for Veo downloads)
+response.usage                                     // credits on Runway; the other three report none
+response.notices                                   // Veo raiMediaFilteredReasons → filtered, xAI respect_moderation → moderated
 yield* response.video.materialize()                // pull bytes before the URL expires

 // Explicit generation control.
 const generation = yield* Video.start(request)     // Generation<VideoResponse>
-generation.id; generation.status; generation.progress; generation.token   // token is serializable JSON
+generation.id; generation.status; generation.progress; generation.position; generation.token
 yield* generation.await({ poll })                  // VideoResponse
-yield* generation.cancel()
+yield* generation.cancel()                         // fal PUT cancel_url, Runway DELETE /tasks/{id}; no-op for Veo and xAI

-// Resume from another process.
-const resumed = yield* Video.resume(model, token)  // Generation<VideoResponse>
+// Resume from another process. The token is validated against the route's codec and refreshed once.
+const resumed = yield* Video.resume(model, JSON.parse(saved))

 // Progress as a stream.
-yield* Video.stream(request)                       // Stream<VideoEvent>: generation-queued { position } | generation-progress { progress, logs } | video { index, video } | finish
+yield* Video.stream(request, { poll })             // Stream<VideoEvent>: generation-queued { id, position } | generation-progress { id, progress } | video { index, video } | finish { usage, notices }
 ```

-Webhooks: `Video.complete(model, token, webhook)` finishes a generation from a webhook payload without polling. Token shape is route-owned and opaque (Veo operation name, fal `response_url`, Runway task id).
+Tokens are route-owned JSON: Veo `{ operation }`, xAI `{ requestID }`, Runway `{ taskID }`, fal
+`{ requestID, statusURL, responseURL, cancelURL }` (fal's follow-up URLs are authoritative and absolute). Common-field
+lowering per provider: Veo takes inline media only and rejects `audio: false` and `n > 1`; xAI rejects `seed` and
+`negativePrompt` and routes a `video` input to edits or (`providerOptions.mode: "extend"`) extensions; fal rejects
+`durationSeconds`, `references`, and `frames.last` because the field names and enums differ per model; Runway passes
+`aspectRatio` through as its pixel `ratio` and rejects `n`.
+
+Deferred: `Video.complete(model, token, webhook)` (finish from a webhook payload without polling) and provider poll
+hints (none of the four providers emit one). Later providers: Luma, Kling, MiniMax, Replicate.

 #### Speech (TTS)

+Shipped in phase 3 (`src/speech.ts`, `src/speech-client.ts`, protocols `openai-speech`, `google-speech`,
+`elevenlabs-speech`, `cartesia-speech`, `deepgram-speech`; new `ElevenLabs`, `Cartesia`, and `Deepgram` facades).
+
 ```ts
 const request = Speech.request({
-  model: elevenlabs.speech("eleven_v3"),
+  model: elevenlabs.speech("eleven_flash_v2_5"),
  text: "Hello from OpenCode.",
-  voice: "JBFqnCBsd6RMkjVDRZzb",                  // name, uuid, or { id } — provider-normalized
+  voice: "JBFqnCBsd6RMkjVDRZzb",                  // provider-native identifier, or { id }
  format: "mp3",                                   // mp3 | wav | pcm | opus | aac | flac | (string & {})
  speed: 1.0,
  language: "en",
-  instructions: "Warm, unhurried.",
-  providerOptions: { stability: 0.5 },
+  instructions: "Warm, unhurried.",                // only OpenAI; elsewhere fails typed
+  timestamps: true,                                // request alignment; routes without it fail typed
+  providerOptions: { voice_settings: { stability: 0.5 } },
 })

-const response = yield* Speech.generate(request)   // SpeechResponse: audio: Media.Asset, timestamps?, usage
-yield* Speech.stream(request)                      // Stream<SpeechEvent>: audio-delta { chunk } | timestamps { words } | finish
+const response = yield* Speech.generate(request)   // SpeechResponse: audio: Media.Asset, timestamps?, usage?, providerMetadata?
+yield* Speech.stream(request)                      // Stream<SpeechEvent>: audio-delta { chunk } | timestamps { items } | finish { audio, usage? }
 ```

-Streaming TTS is first-class on day one: OpenAI `stream_format: sse`, ElevenLabs `/stream`, Cartesia SSE, Deepgram chunked. Input-streaming TTS (WS, text arrives incrementally) is a later `Speech.session(...)` scoped resource, not part of `generate`.
+Execution is `MediaProtocol.stream` for every provider: one request whose body is framed and folded by a `step`
+state machine, with `generate` running the same stream and collecting it. The route submits the request with its
+`mode` (`"generate" | "stream"`), which lets one provider stay one protocol — OpenAI adds `stream_format: "sse"` (except `tts-1`/`tts-1-hd`, which stream raw bytes), ElevenLabs appends
+`/stream`, Cartesia switches `/tts/bytes` to `/tts/sse`, Gemini switches `generateContent` to
+`streamGenerateContent`. The terminal `finish` event carries the assembled asset (every provider's stream is
+concatenable chunks), so stream consumers also get the whole file and `generate` is just "take `finish`, gather
+`timestamps`". The cost is memory: a stream holds every chunk until `finish`, so even a consumer that only plays deltas
+keeps the whole clip in memory. That is bounded by the providers' input text limits (a few minutes of audio); a
+long-form or session API would need an opt-out.
+
+**Voice.** `voice?: string | { id: string }`. A string is passed through as the provider's native identifier — a
+name on OpenAI and Gemini, a voice id on ElevenLabs (path segment) and Cartesia. `{ id }` selects an OpenAI custom
+voice and is treated as the plain string on routes that do not distinguish custom from built-in. Deepgram's voice is
+the model id (`aura-2-thalia-en`), so `voice` is `unsupported` there. There is no cross-provider voice catalog or
+name→id resolution. Multi-speaker (Gemini `speechConfig.multiSpeakerVoiceConfig`) and per-voice settings
+(ElevenLabs `voice_settings`) go through `providerOptions`.
+
+**Format and PCM.** `format` is container-level; provider sample rates and bitrates live under `providerOptions`
+(ElevenLabs `outputFormat`, Cartesia `sampleRate`/`bitRate`/`encoding`, Deepgram `encoding`/`container`/`sampleRate`/
+`bitRate`). Each protocol maps `format` to its native value (ElevenLabs `mp3_44100_128`/`pcm_24000`/`wav_24000`/
+`opus_48000_64`, Cartesia `{ container, encoding, sample_rate }`, Deepgram `encoding`+`container`) and declares the
+asset's media type rather than sniffing, because headerless PCM can look like an MPEG frame sync. Headerless PCM
+always carries `info.encoding`, `info.sampleRate`, and `info.channels`; its media type is the provider's declaration
+(Gemini `audio/L16;codec=pcm;rate=24000`, Deepgram's `content-type`) or `audio/pcm`. Gemini returns PCM only, so any
+other `format` is rejected rather than wrapped as WAV by the route. Every `format` value a route cannot produce (unknown
+to it, a container on Cartesia SSE, WAV on an ElevenLabs stream, anything but PCM on Gemini) fails the same way as an
+unsupported field: `UnsupportedOperation` with `operation: "media.format"`.
+
+**Timestamps.** `timestamps: true` on the request asks for alignment. ElevenLabs selects the `with-timestamps`
+endpoints (character-level, NDJSON when streaming); Cartesia sets `add_timestamps` on `/tts/sse` (word-level; a
+`generate` with timestamps collects the SSE stream). OpenAI, Gemini, and Deepgram reject it.
+
+Common-field lowering per provider:
+
+| Provider | `voice` | `speed` | `language` | `instructions` | `timestamps` | Usage |
+|---|---|---|---|---|---|---|
+| OpenAI | `voice` (name or `{ id }`) | `speed` | unsupported | `instructions` | unsupported | `tokens` from SSE `speech.audio.done` only |
+| Gemini | `prebuiltVoiceConfig.voiceName` | unsupported | `speechConfig.languageCode` | unsupported (direct in text) | unsupported | `tokens` from `usageMetadata` |
+| ElevenLabs | path voice id (required) | `voice_settings.speed` | `language_code` | unsupported | `with-timestamps` | `credits` from `character-cost` header |
+| Cartesia | `voice` (required) | `generation_config.speed` | `language` | unsupported | `add_timestamps` | none |
+| Deepgram | unsupported (voice is the model) | `speed` query | unsupported | unsupported | unsupported | `characters` from `dg-char-count` header |
+
+Deferred: `Speech.session(...)` — input-streaming TTS where text arrives incrementally over a WebSocket (ElevenLabs
+`stream-input`, Cartesia WebSocket contexts, Deepgram WebSocket speak) — is a separate scoped resource, not part of
+`generate`/`stream`, and ships with the realtime work in phase 5.

 #### Transcription (STT)

+Shipped as the second half of phase 3 (`src/transcription.ts`, `src/transcription-client.ts`, protocols
+`openai-transcription`, `google-transcription`, `deepgram-transcription`, `assemblyai-transcription`; new `AssemblyAI`
+facade).
+
 ```ts
 const request = Transcription.request({
-  model: openai.transcription("gpt-4o-transcribe"),
-  audio: Media.file("./call.wav"),
-  language: "en",
-  prompt: "Names: Shoubhit, OpenCode.",
-  timestamps: "word",                              // none | segment | word
+  model: openai.transcription("gpt-4o-transcribe-diarize"),
+  audio: yield* Media.file("./call.wav"),
+  language: "en",                                  // provider-native passthrough
+  timestamps: "segment",                           // none | segment | word
  diarize: true,
-  providerOptions: { chunkingStrategy: "auto" },
+  speakers: 2,                                     // expected count, hint only (AssemblyAI)
+  providerOptions: { known_speaker_names: ["agent"] },
 })

 const response = yield* Transcription.generate(request)
-response.text; response.segments; response.words; response.language; response.durationSeconds
-yield* Transcription.stream(request)               // Stream<TranscriptionEvent>: text-delta | segment | finish
+response.text; response.segments; response.words; response.language; response.durationSeconds; response.usage
+yield* Transcription.stream(request)               // Stream<TranscriptionEvent>: generation-queued | generation-progress | text-delta | segment | finish
+const generation = yield* Transcription.start(request) // queued routes only
+yield* Transcription.resume(model, token)
 ```

-Realtime STT over WebSocket is the same future `session` shape as input-streaming TTS.
+Transcription is the first modality whose providers span all three protocol kinds, and it needed no fourth kind.
+Every `MediaRoute` now carries its `kind`; `TranscriptionRoute` is the union of the inline, stream, and queued routes;
+`TranscriptionModel.fromRoute` is overloaded per protocol kind (arity picks the overload: `<Options>`,
+`<Options, Frame, State>`, `<Options, Token>`) and composes through `MediaRoute.inline` / `stream` / `queued`; and
+`TranscriptionClient` dispatches on `route.kind`. `generate` on a queued route is `start` then `await`; `stream` on an
+inline route is the response as a single `finish`, and on a queued route it is the status observations followed by
+`finish`. `start` / `resume` on a non-queued route fail with `UnsupportedOperation` (`transcription.start`). The
+`finish` event carries the whole transcript (text, segments, words, language, duration, usage), so the stream route's
+`collect` is just "take `finish`".
+
+The route layer gained a `binary` body with array-valued `query` (Deepgram) and `Queued.start.prepare` (AssemblyAI's
+upload); `packages/ai/AGENTS.md` (Media Routes) describes both.
+
+Settled rules:
+
+- **Timestamps.** A granularity the selected route or model cannot produce fails as `UnsupportedOperation`
+  (`media.timestamps`), following Speech; a route that returns more than asked (Deepgram and AssemblyAI always return
+  words) is not stripped. Segments always carry start and end times: Gemini times each transcription part from its
+  word offsets, so segment timestamps and diarization also request word offsets there.
+- **Diarization.** `diarize` means segments (and words, where the provider labels them) carry `speaker`. Labels are
+  provider-native strings — OpenAI `A` or a known speaker name, Deepgram `0`, Gemini `spk:0`, AssemblyAI `A` — with no
+  cross-provider speaker model. `speakers` is a hint; only AssemblyAI (`speakers_expected`) accepts it.
+- **Language** is passed through (`language`, OpenAI `gpt-transcribe` `languages[]`, Gemini `languageCodes`,
+  AssemblyAI `language_code`). `response.language` is the provider's own value, lowercased but not normalized: an
+  ISO code on most routes, `english` from whisper-1, `en_us` from AssemblyAI. Deepgram and AssemblyAI assume English
+  unless asked to detect, so a missing `language` enables their detection.
+- **Gemini** requires a transcribe model; other model ids fail with `UnsupportedOperation` before the call, because
+  general models ignore `audioTranscriptionConfig` and answer conversationally. Streamed chunks carry whole speaker
+  turns (one part per turn), which join with a space.
+- **Streaming inline providers** emit only `finish`; deltas are never faked.
+- **Units.** AssemblyAI milliseconds and Gemini protobuf durations (`"0.400s"`) are normalized to seconds at the
+  protocol boundary.
+
+| Provider | Kind | Audio input | `timestamps` | `diarize` | Unsupported | Usage |
+|---|---|---|---|---|---|---|
+| OpenAI | stream (`stream: true` in `stream` mode) | multipart `file` (inline only) | `whisper-1` (`verbose_json`); diarize model: `segment` | `gpt-4o-transcribe-diarize` (`diarized_json`) | `speakers`; `prompt` on the diarize model; streaming on `whisper-1` | `tokens` or `seconds` |
+| Gemini | stream (`generateContent` / `streamGenerateContent`) | `inlineData` or Gemini Files `fileData` | `audioTranscriptionConfig.wordTimestamp` | `audioTranscriptionConfig.diarization` | `prompt`, `speakers` | `tokens` |
+| Deepgram | inline | raw body, or JSON `{ url }` | words always; `segment` → `utterances` | `diarize_model=latest` + `utterances` | `prompt`, `speakers` | `seconds` (`metadata.duration`) |
+| AssemblyAI | queued (upload → submit → poll) | `/v2/upload` then `audio_url`, or a URL | words always; `segment` → `speaker_labels` | `speaker_labels` | — | `seconds` (`audio_duration`) |
+
+Deferred: `Transcription.session(...)` — realtime STT over WebSocket (Deepgram live, AssemblyAI streaming, ElevenLabs
+realtime, OpenAI realtime transcription) — is the same future scoped `session` shape as input-streaming TTS and ships
+with the realtime work in phase 5. ElevenLabs Scribe is not implemented yet.

 ### `Generation` — shared async execution

 ```ts
 class Generation<Response> {
  readonly id: string
-  readonly model: MediaModel
+  readonly route: GenerationRoute<Response>        // token-free: { status, result, cancel?: Effect; pollHint? } closed over the decoded token
  readonly token: unknown                          // route-owned serializable JSON
  readonly status: "queued" | "running" | "completed" | "failed" | "cancelled" | "expired"
  readonly progress?: number                       // 0..1, normalized
  readonly position?: number
  readonly expiresAt?: number
  refresh(): Effect<Generation<Response>, AIError>
-  await(options?: { poll?: Poll }): Effect<Response, AIError>
+  result(): Effect<Response, AIError>
+  await(options?: AwaitOptions): Effect<Response, AIError>
  cancel(): Effect<void, AIError>
-  events(options?): Stream<GenerationEvent, AIError>
+  events(options?: AwaitOptions): Stream<GenerationEvent, AIError>   // fails with Timeout past poll.timeout, checked per observation
 }

+AwaitOptions = { poll?: Poll }
 Poll = { interval?: Duration; timeout?: Duration; schedule?: Schedule }   // route may override from provider hints (`openai-poll-after-ms`)
 ```

@@ -273,11 +386,14 @@ Existing facades gain per-modality selectors; the modality routes each facade pr
 | Facade | llm | image | video | speech | transcription | other |
 |---|---|---|---|---|---|---|
 | `OpenAI` | responses (default), chat | Images API | Sora (deprecated 2026-09-24) | ✓ | ✓ | |
-| `Google` | Gemini | Gemini-native (default), `imagen` | Veo | Gemini TTS | Gemini transcribe | |
+| `Google` | Gemini | Gemini-native (default), `imagen` | Veo | Gemini TTS | `gemini-3.5-transcribe` | |
 | `XAI` | ✓ | ✓ | ✓ | | | |
 | `ElevenLabs` | | | | ✓ | Scribe | soundEffect, music |
+| `Cartesia` | | | | ✓ | | |
+| `Deepgram` | | | | Aura | ✓ | |
 | `Fal` | | ✓ | ✓ | | | |
-| `Replicate`, `Runway`, `Luma`, `Kling`, `MiniMax`, `Deepgram`, `Cartesia`, `AssemblyAI`, `BlackForestLabs`, `Stability` | | per provider | | | | |
+| `AssemblyAI` | | | | | ✓ (queued) | |
+| `Replicate`, `Runway`, `Luma`, `Kling`, `MiniMax`, `BlackForestLabs`, `Stability` | | per provider | | | | |

 New facades follow the existing one-file-per-provider rule. Package entrypoints are modality-specific, such as `@opencode/ai/providers/openai/images`, and return the concrete model.

@@ -288,10 +404,10 @@ New facades follow the existing one-file-per-provider rule. Package entrypoints
 Media does not fit the LLM four-axis route (SSE frames → event state machine) except for streaming TTS/STT. Reuse `Endpoint`, `Auth`, `Framing`, `RequestExecutor`, and add media protocol kinds:

 - `MediaProtocol.inline` — `body.from(request)` (JSON, multipart, or query), `response.decode(response)` (JSON, or binary body → `Media.Asset`).
- `MediaProtocol.queued` — `start`, `status`, `result`, `cancel`, optional `download`, `pollHint`, `token` schema.
- `MediaProtocol.stream` — framing + `step` state machine emitting modality events, same discipline as LLM protocols.
+- `MediaProtocol.queued` — `start` (body + decode to `{ token, snapshot }`), `status`, `result`, optional `cancel`, `pollHint`, and a `token` codec. `result` is always a separate GET (against the status document for Veo/xAI/Runway, fal's `response_url` otherwise) so `await` after `start` and after `resume` share one path. `PollContext.auth` hands the auth headers the route sent to the protocol for output URLs that need them (Veo downloads); they become transient `Media.Asset.headers`, never part of `source`. There is no separate `download` step: `Media.Asset.bytes()` downloads through the executor with those headers. `MediaRoute.inline(...)` / `MediaRoute.queued(...)` compose each kind with endpoint and auth; the queued route decodes the token once and hands `Generation` a token-free `{ status, result, cancel? }`.
+- `MediaProtocol.stream` — `body.from(request)` over the request plus its `mode`, `frames` (a function that picks the framing for the call: `Framing.sse`, `lines`, `document`, or the raw bytes), fresh per-response `initial()` state, `step` emitting modality events, and `finish(state, context)` — with the observed response for header-only usage — emitting exactly one terminal event or failing as an incomplete stream. The route fills `reason.http` on stream errors. `MediaRoute.stream(...)` exposes `stream` and `generate` (the same stream folded by the modality's `collect`).

-`Route.make` for media composes one protocol kind with endpoint/auth. The existing `ImageRoute { generate(request, execute) }` is the ad-hoc version of `inline` and gets folded in.
+`MediaRoute.inline` / `MediaRoute.queued` / `MediaRoute.stream` compose one protocol kind with endpoint/auth and tag the route with its `kind`; `ImageModel`/`VideoModel`/`SpeechModel`/`TranscriptionModel` share the `MediaModel` base (`src/media-model.ts`).

 ### LLM integration

@@ -320,8 +436,8 @@ Foundation + Image ship together as the reference implementation, serially. Vide
 ## Phasing

 1. **Foundation** — per-modality selectors, `Media`, `Generation`, `Poll`, `Usage` union, `MediaProtocol` kinds, `@opencode/ai/promise` with `llm` + `image`. Port the five existing image protocols onto it. Unify `MediaPart` and add the `media` LLM event (fixes Gemini image output being dropped).
-2. **Video** — Veo, xAI, fal, Runway first. Then Luma, Kling, MiniMax, Replicate.
-3. **Speech + Transcription** — OpenAI, ElevenLabs, Gemini TTS, Deepgram, Cartesia, AssemblyAI. Streaming TTS from the start.
+2. **Video** — ✅ Veo, xAI, fal, Runway shipped (`MediaProtocol.queued`, `Video.start/generate/resume/stream`, promise `ai.video`). Deferred: `Video.complete` (webhooks), Luma, Kling, MiniMax, Replicate.
+3. **Speech + Transcription** — ✅ Speech: OpenAI, Gemini TTS, ElevenLabs, Cartesia, Deepgram shipped (`MediaProtocol.stream`, `Speech.generate/stream`, promise `ai.speech`). ✅ Transcription: OpenAI, Gemini, Deepgram, AssemblyAI shipped across all three route kinds (`Transcription.generate/stream/start/resume`, promise `ai.transcription`). Pending: ElevenLabs Scribe. Deferred: `Speech.session` and `Transcription.session` (WebSocket streaming).
 4. **Image queued routes and partials** — BFL, fal, Replicate, Stability; OpenAI `partial_images` streaming.
 5. **Later** — ElevenLabs music/SFX, Lyria, `Speech.session` / `Transcription.session`, realtime.

@@ -1,6 +1,6 @@
 {
  "$schema": "https://json.schemastore.org/package.json",
-  "version": "2.0.14",
+  "version": "2.0.15",
  "name": "@opencode/ai",
  "type": "module",
  "license": "MIT",
@@ -104,6 +104,83 @@ const PROVIDERS: ReadonlyArray<Provider> = [
    vars: [{ name: "XAI_API_KEY" }],
    validate: (env) => validateBearer("https://api.x.ai/v1/models", Redacted.make(env.XAI_API_KEY)),
  },
+  {
+    id: "fal",
+    label: "fal",
+    tier: "canary",
+    note: "fal queue video recorded tests",
+    vars: [{ name: "FAL_KEY" }],
+    // fal has no free authenticated list endpoint; a 404 for an unknown request id proves the key was accepted.
+    validate: (env) =>
+      Effect.gen(function* () {
+        const http = yield* HttpClient.HttpClient
+        const response = yield* http.execute(
+          HttpClientRequest.get(
+            "https://queue.fal.run/fal-ai/veo3.1/requests/00000000-0000-0000-0000-000000000000/status",
+          ).pipe(HttpClientRequest.setHeaders({ authorization: `Key ${Redacted.value(Redacted.make(env.FAL_KEY))}` })),
+        )
+        if (response.status === 404) return undefined
+        return yield* responseError(response)
+      }),
+  },
+  {
+    id: "runway",
+    label: "Runway",
+    tier: "canary",
+    note: "Runway task video recorded tests",
+    vars: [{ name: "RUNWAYML_API_SECRET" }],
+    validate: (env) =>
+      validateBearer("https://api.dev.runwayml.com/v1/organization", Redacted.make(env.RUNWAYML_API_SECRET), {
+        "X-Runway-Version": "2024-11-06",
+      }),
+  },
+  {
+    id: "elevenlabs",
+    label: "ElevenLabs",
+    tier: "canary",
+    note: "ElevenLabs text-to-speech recorded tests",
+    vars: [{ name: "ELEVENLABS_API_KEY" }],
+    validate: (env) =>
+      HttpClientRequest.get("https://api.elevenlabs.io/v1/models").pipe(
+        HttpClientRequest.setHeader("xi-api-key", Redacted.value(Redacted.make(env.ELEVENLABS_API_KEY))),
+        executeRequest,
+      ),
+  },
+  {
+    id: "cartesia",
+    label: "Cartesia",
+    tier: "canary",
+    note: "Cartesia text-to-speech recorded tests",
+    vars: [{ name: "CARTESIA_API_KEY" }],
+    validate: (env) =>
+      validateBearer("https://api.cartesia.ai/voices?limit=1", Redacted.make(env.CARTESIA_API_KEY), {
+        "Cartesia-Version": "2026-08-14",
+      }),
+  },
+  {
+    id: "deepgram",
+    label: "Deepgram",
+    tier: "canary",
+    note: "Deepgram Aura text-to-speech and Nova transcription recorded tests",
+    vars: [{ name: "DEEPGRAM_API_KEY" }],
+    validate: (env) =>
+      HttpClientRequest.get("https://api.deepgram.com/v1/projects").pipe(
+        HttpClientRequest.setHeader("authorization", `Token ${Redacted.value(Redacted.make(env.DEEPGRAM_API_KEY))}`),
+        executeRequest,
+      ),
+  },
+  {
+    id: "assemblyai",
+    label: "AssemblyAI",
+    tier: "canary",
+    note: "AssemblyAI queued transcription recorded tests",
+    vars: [{ name: "ASSEMBLYAI_API_KEY" }],
+    validate: (env) =>
+      HttpClientRequest.get("https://api.assemblyai.com/v2/transcript?limit=1").pipe(
+        HttpClientRequest.setHeader("authorization", Redacted.value(Redacted.make(env.ASSEMBLYAI_API_KEY))),
+        executeRequest,
+      ),
+  },
  {
    id: "cloudflare-ai-gateway",
    label: "Cloudflare AI Gateway",
@@ -1,4 +1,4 @@
-import { Duration, Effect, Schedule, Schema, Stream } from "effect"
+import { Clock, Duration, Effect, Schedule, Schema, Stream } from "effect"
 import { AIError, TimeoutError } from "./schema/errors.js"

 export const Status = Schema.Literals(["queued", "running", "completed", "failed", "cancelled", "expired"])
@@ -15,13 +15,13 @@ export interface Snapshot {
 }

 /**
- * Route-owned generation operations. `token` is the route's serializable handle (operation name, task id, response URL)
- * so a generation can be resumed from another process; its shape is opaque to `Generation`.
+ * Route-owned generation operations for one generation. The media route decodes its serializable token once (from the
+ * submission response or a `resume` input) and closes over it, so `Generation` never sees the token's shape.
 */
 export interface Route<Response> {
-  readonly status: (token: unknown) => Effect.Effect<Snapshot, AIError>
-  readonly result: (token: unknown) => Effect.Effect<Response, AIError>
-  readonly cancel?: (token: unknown) => Effect.Effect<void, AIError>
+  readonly status: Effect.Effect<Snapshot, AIError>
+  readonly result: Effect.Effect<Response, AIError>
+  readonly cancel?: Effect.Effect<void, AIError>
  /** Provider polling hint (e.g. `openai-poll-after-ms`) that overrides the default interval for the next poll. */
  readonly pollHint?: (snapshot: Snapshot) => Duration.Duration | undefined
 }
@@ -33,13 +33,28 @@ export interface Poll {
  readonly schedule?: Schedule.Schedule<unknown, Snapshot>
 }

+export interface AwaitOptions {
+  readonly poll?: Poll
+}
+
 export const DEFAULT_POLL_INTERVAL = Duration.seconds(5)
 export const DEFAULT_POLL_TIMEOUT = Duration.minutes(10)

-export type Event =
-  | { readonly type: "generation-queued"; readonly id: string; readonly position?: number }
-  | { readonly type: "generation-progress"; readonly id: string; readonly progress?: number }
-  | { readonly type: "generation-finished"; readonly id: string; readonly status: Status }
+export const QueuedEvent = Schema.Struct({
+  type: Schema.tag("generation-queued"),
+  id: Schema.String,
+  position: Schema.optional(Schema.Number),
+}).annotate({ identifier: "Generation.Event.Queued" })
+
+export const ProgressEvent = Schema.Struct({
+  type: Schema.tag("generation-progress"),
+  id: Schema.String,
+  progress: Schema.optional(Schema.Number),
+}).annotate({ identifier: "Generation.Event.Progress" })
+
+export type Observation = Schema.Schema.Type<typeof QueuedEvent> | Schema.Schema.Type<typeof ProgressEvent>
+
+export type Event = Observation | { readonly type: "generation-finished"; readonly id: string; readonly status: Status }

 const TERMINAL: ReadonlySet<Status> = new Set(["completed", "failed", "cancelled", "expired"])

@@ -52,6 +67,7 @@ export class Generation<Response> {

  constructor(
    readonly route: Route<Response>,
+    /** Route-owned serializable JSON; pass it to the modality's `resume` from another process. */
    readonly token: unknown,
    snapshot: Snapshot,
  ) {
@@ -77,51 +93,79 @@ export class Generation<Response> {
  }

  refresh(): Effect.Effect<Generation<Response>, AIError> {
-    return this.route.status(this.token).pipe(Effect.map((snapshot) => new Generation(this.route, this.token, snapshot)))
+    return this.route.status.pipe(Effect.map((snapshot) => new Generation(this.route, this.token, snapshot)))
+  }
+
+  /** Fetch the result without polling; non-completed terminal generations fail with the provider's terminal body. */
+  result(): Effect.Effect<Response, AIError> {
+    return this.route.result
  }

  /** Poll until the generation reaches a terminal status, then fetch the result. Fails with a `Timeout` reason on deadline. */
-  await(options?: { readonly poll?: Poll }): Effect.Effect<Response, AIError> {
+  await(options?: AwaitOptions): Effect.Effect<Response, AIError> {
    const timeout = Duration.fromInputUnsafe(options?.poll?.timeout ?? DEFAULT_POLL_TIMEOUT)
    const settled = this.terminal ? Effect.succeed(this) : this.poll(options?.poll)
    return settled.pipe(
      // Non-completed terminal states also go through `result` so the route can surface its provider failure body.
-      Effect.flatMap((generation) => generation.route.result(generation.token)),
-      Effect.timeoutOrElse({
-        duration: timeout,
-        orElse: () =>
-          new AIError({
-            reason: new TimeoutError({
-              message: `Generation ${this.id} did not finish within ${Duration.format(timeout)}`,
-              timeoutMs: Duration.toMillis(timeout),
-            }),
-          }),
-      }),
+      Effect.flatMap((generation) => generation.result()),
+      Effect.timeoutOrElse({ duration: timeout, orElse: () => this.timeoutError(timeout) }),
    )
  }

  cancel(): Effect.Effect<void, AIError> {
-    return this.route.cancel?.(this.token) ?? Effect.void
+    return this.route.cancel ?? Effect.void
  }

-  /** Status observations as a stream, ending after the first terminal observation. */
-  events(options?: { readonly poll?: Poll }): Stream.Stream<Event, AIError> {
-    const observations = this.terminal
-      ? Stream.make(this)
-      : Stream.fromEffectSchedule(this.refresh(), this.schedule(options?.poll)).pipe(
-          Stream.takeUntil((generation) => generation.terminal),
-        )
-    return observations.pipe(
-      Stream.map((generation): Event => {
-        if (generation.terminal) return { type: "generation-finished", id: generation.id, status: generation.status }
-        if (generation.status === "queued") return { type: "generation-queued", id: generation.id, position: generation.position }
-        return { type: "generation-progress", id: generation.id, progress: generation.progress }
-      }),
+  /**
+   * Status observations as a stream, ending after the first terminal observation. Each poll is bounded by the time
+   * remaining until `poll.timeout`, so a hung status request fails the stream instead of stalling it. (`Stream.interruptWhen`
+   * would express this directly but deadlocks under `TestClock` when the source completes while the timer sleeps.)
+   */
+  events(options?: AwaitOptions): Stream.Stream<Event, AIError> {
+    if (this.terminal) return Stream.make(this.event())
+    const timeout = Duration.fromInputUnsafe(options?.poll?.timeout ?? DEFAULT_POLL_TIMEOUT)
+    return Stream.unwrap(
+      Clock.currentTimeMillis.pipe(
+        Effect.map((start) => {
+          const deadline = start + Duration.toMillis(timeout)
+          const refresh = Clock.currentTimeMillis.pipe(
+            Effect.flatMap((now) =>
+              this.refresh().pipe(
+                Effect.timeoutOrElse({
+                  duration: Duration.millis(Math.max(0, deadline - now)),
+                  orElse: () => this.timeoutError(timeout),
+                }),
+              ),
+            ),
+          )
+          return Stream.fromEffectSchedule(refresh, this.schedule(options?.poll)).pipe(
+            Stream.takeUntil((generation) => generation.terminal),
+            Stream.map((generation) => generation.event()),
+          )
+        }),
+      ),
    )
  }

+  private event(): Event {
+    if (this.terminal) return { type: "generation-finished", id: this.id, status: this.status }
+    if (this.status === "queued") return { type: "generation-queued", id: this.id, position: this.position }
+    return { type: "generation-progress", id: this.id, progress: this.progress }
+  }
+
+  private timeoutError(timeout: Duration.Duration) {
+    return new AIError({
+      reason: new TimeoutError({
+        message: `Generation ${this.id} did not finish within ${Duration.format(timeout)}`,
+        timeoutMs: Duration.toMillis(timeout),
+      }),
+    })
+  }
+
  private poll(poll: Poll | undefined) {
-    return this.refresh().pipe(Effect.repeat({ schedule: this.schedule(poll), until: (generation) => generation.terminal }))
+    return this.refresh().pipe(
+      Effect.repeat({ schedule: this.schedule(poll), until: (generation) => generation.terminal }),
+    )
  }

  private schedule(poll: Poll | undefined): Schedule.Schedule<unknown, Generation<Response>> {
@@ -135,3 +179,13 @@ export class Generation<Response> {
    )
  }
 }
+
+export const resultEvents = <Response, A>(
+  generation: Generation<Response>,
+  expand: (response: Response) => ReadonlyArray<A>,
+  options?: AwaitOptions,
+): Stream.Stream<Observation | A, AIError> =>
+  generation.events(options).pipe(
+    Stream.filter((event): event is Observation => event.type !== "generation-finished"),
+    Stream.concat(Stream.fromIterableEffect(Effect.map(generation.result(), expand))),
+  )
@@ -48,7 +48,7 @@ export const layer: Layer.Layer<Service, never, RequestExecutor.Service> = Layer
      generate,
      // Inline routes have no partial frames yet; the stream is the completed response expanded into events.
      stream: (request) =>
-        Stream.unwrap(generate(request).pipe(Effect.map((response) => Stream.fromIterable(responseEvents(response))))),
+        Stream.fromIterableEffect(Effect.map(generate(request), responseEvents)),
    })
  }),
 )
@@ -1,17 +1,9 @@
 import { Effect, Schema, Stream } from "effect"
 import { Media } from "./media.js"
-import { Endpoint } from "./route/endpoint.js"
+import { MediaModel, composeRoute, tryRequest } from "./media-model.js"
 import { MediaRoute } from "./route/media.js"
 import type { MediaProtocol } from "./route/media-protocol.js"
-import {
-  AIError,
-  HttpOptions,
-  InvalidRequestError,
-  MediaUsage,
-  ModelID,
-  ProviderID,
-  ProviderMetadata,
-} from "./schema/index.js"
+import { AIError, HttpOptions, MediaUsage, ProviderMetadata } from "./schema/index.js"
 import { ImageClient, Service } from "./image-client.js"

 // ---------------------------------------------------------------------------
@@ -25,27 +17,11 @@ export type ImageRoute<Options extends ImageOptions = ImageOptions> = MediaRoute
  ImageResponse
 >

-export class ImageModel<Options extends ImageOptions = ImageOptions> {
-  declare protected readonly _Options: (options: Options) => Options
-  readonly id: ModelID
-  readonly provider: ProviderID
-  readonly route: ImageRoute<Options>
-  readonly http?: HttpOptions
+export class ImageModel<Options extends ImageOptions = ImageOptions> extends MediaModel<ImageRoute<Options>, Options> {
+  declare protected readonly _ImageModel: void

-  constructor(input: ImageModel.Input<Options>) {
-    this.id = input.id
-    this.provider = input.provider
-    this.route = input.route
-    this.http = input.http
-  }
-
-  static make<Options extends ImageOptions = ImageOptions>(input: ImageModel.MakeInput<Options>) {
-    return new ImageModel<Options>({
-      id: ModelID.make(input.id),
-      provider: ProviderID.make(input.provider),
-      route: input.route,
-      http: input.http,
-    })
+  static make<Options extends ImageOptions = ImageOptions>(input: MediaModel.Input<ImageRoute<Options>>) {
+    return new ImageModel<Options>(input)
  }

  /** Compose an inline image protocol with its canonical path into a model for one deployment. */
@@ -53,44 +29,20 @@ export class ImageModel<Options extends ImageOptions = ImageOptions> {
    route: ImageModel.RouteInput<Options>,
    input: MediaRoute.ModelInput,
  ) {
-    return ImageModel.make<Options>({
+    return new ImageModel<Options>({
      id: input.id,
      provider: route.provider,
      http: input.http,
-      route: MediaRoute.make({
-        id: route.id,
-        provider: route.provider,
-        protocol: route.protocol,
-        endpoint: Endpoint.path(route.path, { baseURL: input.baseURL ?? route.baseURL }),
-        auth: input.auth,
-        headers: input.headers,
-      }),
+      route: composeRoute(MediaRoute.inline, route, input),
    })
  }
 }

 export namespace ImageModel {
-  export interface Input<Options extends ImageOptions = ImageOptions> {
-    readonly id: ModelID
-    readonly provider: ProviderID
-    readonly route: ImageRoute<Options>
-    readonly http?: HttpOptions
-  }
-
-  export interface MakeInput<Options extends ImageOptions = ImageOptions>
-    extends Omit<Input<Options>, "id" | "provider"> {
-    readonly id: string | ModelID
-    readonly provider: string | ProviderID
-  }
-
-  export interface RouteInput<Options extends ImageOptions = ImageOptions> {
-    readonly id: string
-    readonly provider: string | ProviderID
-    readonly protocol: MediaProtocol.Inline<ImageRequestFor<Options>, ImageResponse>
-    readonly path: Endpoint.EndpointPart<MediaProtocol.Body, ImageRequestFor<Options>>
-    /** Canonical base URL; `ModelInput.baseURL` overrides it per deployment. */
-    readonly baseURL?: string
-  }
+  export type RouteInput<Options extends ImageOptions = ImageOptions> = MediaModel.RouteInput<
+    ImageRequestFor<Options>,
+    MediaProtocol.Inline<ImageRequestFor<Options>, ImageResponse>
+  >
 }

 export const ImageModelSchema = Schema.declare((value): value is ImageModel => value instanceof ImageModel, {
@@ -107,11 +59,8 @@ export const ImageSize = Schema.declare<ImageSize>(
  { title: "ImageSize" },
 )

-export type ImageAspectRatio = `${number}:${number}`
-export const ImageAspectRatio = Schema.declare<ImageAspectRatio>(
-  (value): value is ImageAspectRatio => typeof value === "string" && /^\d+(?:\.\d+)?:\d+(?:\.\d+)?$/.test(value),
-  { title: "ImageAspectRatio" },
-)
+export type ImageAspectRatio = Media.AspectRatio
+export const ImageAspectRatio = Media.AspectRatio

 export type ImageFormat = "png" | "jpeg" | "webp" | (string & {})

@@ -216,17 +165,7 @@ export function request(input: ImageRequest | ImageRequestInput) {
  })
 }

-const requestEffect = (input: ImageRequest | ImageRequestInput) =>
-  Effect.try({
-    try: () => request(input),
-    catch: (error) =>
-      new AIError({
-        reason: new InvalidRequestError({
-          message: error instanceof Error ? error.message : String(error),
-          cause: error,
-        }),
-      }),
-  })
+const requestEffect = (input: ImageRequest | ImageRequestInput) => tryRequest(() => request(input))

 export function generate<const Model extends ImageModel>(
  input: ImageRequestInput<Model>,
@@ -29,9 +29,73 @@ export type {
  ImageRoute,
 } from "./image.js"
 export { Image } from "./image.js"
+export { VideoClient } from "./video-client.js"
+export {
+  VideoAspectRatio,
+  VideoEvent,
+  VideoFrames,
+  VideoModel,
+  VideoModelSchema,
+  VideoRequest,
+  VideoResponse,
+} from "./video.js"
+export type {
+  VideoModelOptions,
+  VideoOptions,
+  VideoRequestFor,
+  VideoRequestInput,
+  VideoResolution,
+  VideoRoute,
+} from "./video.js"
+export { Video } from "./video.js"
+export { SpeechClient } from "./speech-client.js"
+export {
+  SpeechEvent,
+  SpeechModel,
+  SpeechModelSchema,
+  SpeechRequest,
+  SpeechResponse,
+  SpeechTimestamp,
+  SpeechVoice,
+} from "./speech.js"
+export type {
+  SpeechFormat,
+  SpeechModelOptions,
+  SpeechOptions,
+  SpeechRequestFor,
+  SpeechRequestInput,
+  SpeechRoute,
+} from "./speech.js"
+export { Speech } from "./speech.js"
+export { TranscriptionClient } from "./transcription-client.js"
+export {
+  TranscriptionEvent,
+  TranscriptionModel,
+  TranscriptionModelSchema,
+  TranscriptionRequest,
+  TranscriptionResponse,
+  TranscriptionSegment,
+  TranscriptionTimestamps,
+  TranscriptionWord,
+} from "./transcription.js"
+export type {
+  TranscriptionModelOptions,
+  TranscriptionOptions,
+  TranscriptionRequestFor,
+  TranscriptionRequestInput,
+  TranscriptionRoute,
+} from "./transcription.js"
+export { Transcription } from "./transcription.js"
 export { Media } from "./media.js"
 export { Generation } from "./generation.js"
-export type { Event as GenerationEvent, Poll, Route as GenerationRoute, Snapshot as GenerationSnapshot, Status as GenerationStatus } from "./generation.js"
+export type {
+  AwaitOptions as GenerationAwaitOptions,
+  Event as GenerationEvent,
+  Poll,
+  Route as GenerationRoute,
+  Snapshot as GenerationSnapshot,
+  Status as GenerationStatus,
+} from "./generation.js"
 export { Tool, ToolFailure, toDefinitions } from "./tool.js"
 export { ToolRuntime } from "./tool-runtime.js"
 export type { DispatchResult as ToolDispatchResult, ToolSettlement } from "./tool-runtime.js"
@@ -0,0 +1,74 @@
+import { Effect } from "effect"
+import { Endpoint } from "./route/endpoint.js"
+import type { MediaRoute } from "./route/media.js"
+import type { MediaProtocol } from "./route/media-protocol.js"
+import { AIError, HttpOptions, InvalidRequestError, ModelID, ProviderID } from "./schema/index.js"
+
+/**
+ * What every media model carries: ids, the configured route, and deployment `http` overlays. Modality classes
+ * (`ImageModel`, `VideoModel`, `SpeechModel`) extend it with their route type and a nominal marker so one cannot stand
+ * in for the other in requests.
+ */
+export class MediaModel<Route, Options> {
+  declare protected readonly _Options: (options: Options) => Options
+  readonly id: ModelID
+  readonly provider: ProviderID
+  readonly route: Route
+  readonly http?: HttpOptions
+
+  constructor(input: MediaModel.Input<Route>) {
+    this.id = ModelID.make(input.id)
+    this.provider = ProviderID.make(input.provider)
+    this.route = input.route
+    this.http = input.http
+  }
+}
+
+export namespace MediaModel {
+  export interface Input<Route> {
+    readonly id: string | ModelID
+    readonly provider: string | ProviderID
+    readonly route: Route
+    readonly http?: HttpOptions
+  }
+
+  /** A protocol plus its canonical start path; `ModelInput.baseURL` overrides `baseURL` per deployment. */
+  export interface RouteInput<Request extends MediaRoute.MediaRequest, Protocol> {
+    readonly id: string
+    readonly provider: string | ProviderID
+    readonly protocol: Protocol
+    readonly path: Endpoint.EndpointPart<MediaProtocol.Body, Request>
+    readonly baseURL?: string
+    /** Headers the protocol requires on every call, such as a pinned API version; deployment headers win. */
+    readonly headers?: Record<string, string>
+  }
+}
+
+/** Compose a protocol route input with one deployment through `MediaRoute.inline`, `queued`, or `stream`. */
+export const composeRoute = <Request extends MediaRoute.MediaRequest, Protocol, Route>(
+  compose: (input: MediaRoute.Composition<Request> & { readonly protocol: Protocol }) => Route,
+  route: MediaModel.RouteInput<Request, Protocol>,
+  input: MediaRoute.ModelInput,
+): Route =>
+  compose({
+    id: route.id,
+    provider: route.provider,
+    protocol: route.protocol,
+    endpoint: Endpoint.path(route.path, { baseURL: input.baseURL ?? route.baseURL }),
+    auth: input.auth,
+    headers:
+      route.headers === undefined && input.headers === undefined ? undefined : { ...route.headers, ...input.headers },
+  })
+
+/** Lift a synchronous Schema-class constructor into a typed `InvalidRequest` failure. */
+export const tryRequest = <A>(make: () => A): Effect.Effect<A, AIError> =>
+  Effect.try({
+    try: make,
+    catch: (error) =>
+      new AIError({
+        reason: new InvalidRequestError({
+          message: error instanceof Error ? error.message : String(error),
+          cause: error,
+        }),
+      }),
+  })
@@ -34,8 +34,6 @@ const UrlSource = Schema.Struct({
  mediaType: Schema.optional(Schema.String),
  /** Epoch milliseconds after which the provider no longer serves the URL. */
  expiresAt: Schema.optional(Schema.Number),
-  /** Headers required to fetch the URL, such as provider auth for Veo downloads. */
-  headers: Schema.optional(Schema.Record(Schema.String, Schema.String)),
 })

 /** A provider-side handle: OpenAI `file_id`, Gemini file URI, `gs://`, `runway://`, or a prior generation id. */
@@ -55,6 +53,12 @@ export type Source = Schema.Schema.Type<typeof Source>
 // Kind, Info, Notice
 // ---------------------------------------------------------------------------

+export type AspectRatio = `${number}:${number}`
+export const AspectRatio = Schema.declare<AspectRatio>(
+  (value): value is AspectRatio => typeof value === "string" && /^\d+(?:\.\d+)?:\d+(?:\.\d+)?$/.test(value),
+  { title: "Media.AspectRatio" },
+)
+
 export const Kind = Schema.Literals(["image", "video", "audio", "document", "other"])
 export type Kind = Schema.Schema.Type<typeof Kind>

@@ -110,6 +114,8 @@ export class Asset {
  /** Epoch milliseconds after which a `url` source stops resolving. */
  readonly expiresAt?: number
  readonly providerMetadata?: ProviderMetadata
+  /** Transient download credentials for `url` sources; see `Asset.Input.headers`. */
+  readonly headers?: Record<string, string>

  // Derived payload forms are cached on the instance because every protocol lowering re-reads the same payload. The
  // cache is check-then-set (concurrent first reads of a `url` source may both download) and is never observable
@@ -127,6 +133,7 @@ export class Asset {
    this.info = input.info
    this.expiresAt = input.source.type === "url" ? input.source.expiresAt : undefined
    this.providerMetadata = input.providerMetadata
+    this.headers = input.source.type === "url" ? input.headers : undefined
  }

  /** Inline payload without effects, for protocols that embed base64 or data URLs directly. */
@@ -151,7 +158,7 @@ export class Asset {
          ? Effect.fromResult(Encoding.decodeBase64(source.data)).pipe(
              Effect.mapError((cause) => invalid(`Media asset contains invalid base64 data`, cause)),
            )
-          : download(source)
+          : download(source, this.headers)
      return decoded.pipe(Effect.tap((data) => Effect.sync(() => (this.#bytes = data))))
    })
  }
@@ -198,6 +205,12 @@ export namespace Asset {
    readonly source: Source
    readonly info?: Info
    readonly providerMetadata?: ProviderMetadata
+    /**
+     * Headers required to download a `url` source, such as the provider API key Veo demands for its file URIs.
+     * They are runtime-only: never part of `source`, `toJSON()`, or `AssetSchema`, so a persisted asset cannot leak
+     * credentials and cannot be downloaded again after a round-trip. Call `materialize()` before persisting.
+     */
+    readonly headers?: Record<string, string>
  }
 }

@@ -226,10 +239,13 @@ export const AssetSchema = AssetEncoded.pipe(
  }),
 )

-const download = Effect.fn("Media.download")(function* (source: Extract<Source, { readonly type: "url" }>) {
+const download = Effect.fn("Media.download")(function* (
+  source: Extract<Source, { readonly type: "url" }>,
+  headers: Record<string, string> | undefined,
+) {
  const executor = yield* Service
  const response = yield* executor.execute(
-    HttpClientRequest.get(source.url).pipe(HttpClientRequest.setHeaders(source.headers ?? {})),
+    HttpClientRequest.get(source.url).pipe(HttpClientRequest.setHeaders(headers ?? {})),
  )
  const buffer = yield* response.arrayBuffer.pipe(
    Effect.mapError(
@@ -264,8 +280,8 @@ export const url = (
  value: string,
  options?: AssetOptions & Omit<Extract<Source, { readonly type: "url" }>, "type" | "url">,
 ) => {
-  const { mediaType, expiresAt, headers, ...rest } = options ?? {}
-  return from({ type: "url", url: value, mediaType, expiresAt, headers }, rest)
+  const { mediaType, expiresAt, ...rest } = options ?? {}
+  return from({ type: "url", url: value, mediaType, expiresAt }, rest)
 }

 export const ref = (provider: string | ProviderID, id: string, mediaType?: string, options?: AssetOptions) =>
@@ -1,4 +1,5 @@
 import { Effect, Layer, ManagedRuntime, Stream } from "effect"
+import type { AwaitOptions, Generation, Snapshot } from "./generation.js"
 import { Image, ImageModel, ImageRequest, type ImageRequestInput } from "./image.js"
 import { ImageClient } from "./image-client.js"
 import { LLM } from "./index.js"
@@ -6,10 +7,23 @@ import { LLMClient } from "./route/client.js"
 import { RequestExecutor } from "./route/executor.js"
 import { LanguageModel, LLMRequest } from "./schema/index.js"
 import type { RequestInput } from "./llm.js"
+import { Speech, SpeechModel, SpeechRequest, type SpeechRequestInput } from "./speech.js"
+import { SpeechClient } from "./speech-client.js"
+import {
+  Transcription,
+  TranscriptionModel,
+  TranscriptionRequest,
+  type TranscriptionOptions,
+  type TranscriptionRequestInput,
+} from "./transcription.js"
+import { TranscriptionClient } from "./transcription-client.js"
+import { Video, VideoModel, VideoRequest, type VideoOptions, type VideoRequestInput } from "./video.js"
+import { VideoClient } from "./video-client.js"

 /**
- * Promise-first entrypoint for scripts and non-Effect callers. One `ManagedRuntime` hosts the LLM and image clients
- * over a request executor; every method runs the corresponding Effect API and rethrows `AIError` unchanged.
+ * Promise-first entrypoint for scripts and non-Effect callers. One `ManagedRuntime` hosts the LLM, image, video, speech,
+ * and transcription clients over a request executor; every method runs the corresponding Effect API and rethrows
+ * `AIError` unchanged.
 */
 export interface Options {
  /** Executor layer; defaults to `RequestExecutor.fetchLayer`. Inject a recorder or middleware here. */
@@ -23,8 +37,20 @@ export interface RunOptions {
 export type Services =
  | Layer.Success<typeof LLMClient.layer>
  | Layer.Success<typeof ImageClient.layer>
+  | Layer.Success<typeof VideoClient.layer>
+  | Layer.Success<typeof SpeechClient.layer>
+  | Layer.Success<typeof TranscriptionClient.layer>
  | RequestExecutor.Service

+/** Promise view of a `Generation`: its snapshot plus `await`, `refresh`, and `cancel` returning promises. */
+export type GenerationHandle<Response> = Snapshot & {
+  /** Serializable JSON; pass it back to `resume` from another process. */
+  readonly token: unknown
+  readonly await: (options?: AwaitOptions & RunOptions) => Promise<Response>
+  readonly refresh: (options?: RunOptions) => Promise<GenerationHandle<Response>>
+  readonly cancel: (options?: RunOptions) => Promise<void>
+}
+
 const abortEffect = (signal: AbortSignal | undefined) =>
  signal === undefined
    ? Effect.never
@@ -40,9 +66,13 @@ const abortEffect = (signal: AbortSignal | undefined) =>

 export const make = (options: Options = {}) => {
  const runtime = ManagedRuntime.make(
-    Layer.mergeAll(LLMClient.layer, ImageClient.layer).pipe(
-      Layer.provideMerge(options.layer ?? RequestExecutor.fetchLayer),
-    ),
+    Layer.mergeAll(
+      LLMClient.layer,
+      ImageClient.layer,
+      VideoClient.layer,
+      SpeechClient.layer,
+      TranscriptionClient.layer,
+    ).pipe(Layer.provideMerge(options.layer ?? RequestExecutor.fetchLayer)),
  )

  /** Run any package Effect (for example `asset.bytes()`) inside this runtime. */
@@ -61,19 +91,31 @@ export const make = (options: Options = {}) => {
      ),
    )

+  const handle = <Response>(generation: Generation<Response>): GenerationHandle<Response> => ({
+    ...generation.snapshot,
+    token: generation.token,
+    await: (options) => run(generation.await({ poll: options?.poll }), options),
+    refresh: (options) => run(generation.refresh(), options).then(handle),
+    cancel: (options) => run(generation.cancel(), options),
+  })
+
  // The typed `generate`/`stream` overloads take a concrete input or a request, not the union; normalize once here.
  const llmRequest = (input: RequestInput | LLMRequest) => (input instanceof LLMRequest ? input : LLM.request(input))
  const imageRequest = (input: ImageRequestInput | ImageRequest) =>
    input instanceof ImageRequest ? input : Image.request(input)
+  const videoRequest = (input: VideoRequestInput | VideoRequest) =>
+    input instanceof VideoRequest ? input : Video.request(input)
+  const speechRequest = (input: SpeechRequestInput | SpeechRequest) =>
+    input instanceof SpeechRequest ? input : Speech.request(input)
+  const transcriptionRequest = (input: TranscriptionRequestInput | TranscriptionRequest) =>
+    input instanceof TranscriptionRequest ? input : Transcription.request(input)

  return {
    run,
    llm: {
      request: LLM.request,
-      generate: <const Model extends LanguageModel>(
-        input: RequestInput<Model> | LLMRequest,
-        options?: RunOptions,
-      ) => run(LLM.generate(llmRequest(input)), options),
+      generate: <const Model extends LanguageModel>(input: RequestInput<Model> | LLMRequest, options?: RunOptions) =>
+        run(LLM.generate(llmRequest(input)), options),
      stream: <const Model extends LanguageModel>(input: RequestInput<Model> | LLMRequest, options?: RunOptions) =>
        iterate(LLM.stream(llmRequest(input)), options),
    },
@@ -83,10 +125,54 @@ export const make = (options: Options = {}) => {
        input: ImageRequestInput<Model> | ImageRequest,
        options?: RunOptions,
      ) => run(Image.generate(imageRequest(input)), options),
-      stream: <const Model extends ImageModel>(
-        input: ImageRequestInput<Model> | ImageRequest,
+      stream: <const Model extends ImageModel>(input: ImageRequestInput<Model> | ImageRequest, options?: RunOptions) =>
+        iterate(Image.stream(imageRequest(input)), options),
+    },
+    video: {
+      request: Video.request,
+      start: <const Model extends VideoModel>(input: VideoRequestInput<Model> | VideoRequest, options?: RunOptions) =>
+        run(Video.start(videoRequest(input)), options).then(handle),
+      generate: <const Model extends VideoModel>(
+        input: VideoRequestInput<Model> | VideoRequest,
+        options?: AwaitOptions & RunOptions,
+      ) => run(Video.generate(videoRequest(input), { poll: options?.poll }), options),
+      resume: <Options extends VideoOptions>(model: VideoModel<Options>, token: unknown, options?: RunOptions) =>
+        run(Video.resume(model, token), options).then(handle),
+      stream: <const Model extends VideoModel>(
+        input: VideoRequestInput<Model> | VideoRequest,
+        options?: AwaitOptions & RunOptions,
+      ) => iterate(Video.stream(videoRequest(input), { poll: options?.poll }), options),
+    },
+    speech: {
+      request: Speech.request,
+      generate: <const Model extends SpeechModel>(
+        input: SpeechRequestInput<Model> | SpeechRequest,
        options?: RunOptions,
-      ) => iterate(Image.stream(imageRequest(input)), options),
+      ) => run(Speech.generate(speechRequest(input)), options),
+      stream: <const Model extends SpeechModel>(
+        input: SpeechRequestInput<Model> | SpeechRequest,
+        options?: RunOptions,
+      ) => iterate(Speech.stream(speechRequest(input)), options),
+    },
+    transcription: {
+      request: Transcription.request,
+      generate: <const Model extends TranscriptionModel>(
+        input: TranscriptionRequestInput<Model> | TranscriptionRequest,
+        options?: AwaitOptions & RunOptions,
+      ) => run(Transcription.generate(transcriptionRequest(input), { poll: options?.poll }), options),
+      stream: <const Model extends TranscriptionModel>(
+        input: TranscriptionRequestInput<Model> | TranscriptionRequest,
+        options?: AwaitOptions & RunOptions,
+      ) => iterate(Transcription.stream(transcriptionRequest(input), { poll: options?.poll }), options),
+      start: <const Model extends TranscriptionModel>(
+        input: TranscriptionRequestInput<Model> | TranscriptionRequest,
+        options?: RunOptions,
+      ) => run(Transcription.start(transcriptionRequest(input)), options).then(handle),
+      resume: <Options extends TranscriptionOptions>(
+        model: TranscriptionModel<Options>,
+        token: unknown,
+        options?: RunOptions,
+      ) => run(Transcription.resume(model, token), options).then(handle),
    },
    dispose: () => runtime.dispose(),
  }
@@ -0,0 +1,211 @@
+import { Effect, Schema } from "effect"
+import type { HttpClientResponse } from "effect/unstable/http"
+import type { Status } from "../generation.js"
+import { Media } from "../media.js"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
+import { TranscriptionModel, TranscriptionResponse, type TranscriptionRequestFor } from "../transcription.js"
+import { ProviderShared, optionalNull } from "./shared.js"
+import { MediaInput } from "./utils/media-input.js"
+
+const ADAPTER = "assemblyai-transcription"
+const NAME = "AssemblyAI"
+const PROVIDER = ProviderID.make("assemblyai")
+export const DEFAULT_BASE_URL = "https://api.assemblyai.com"
+export const PATH = "/v2/transcript"
+export const UPLOAD_PATH = "/v2/upload"
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+export type AssemblyAITranscriptionOptions = {
+  readonly keyterms_prompt?: ReadonlyArray<string>
+  readonly punctuate?: boolean
+  readonly format_text?: boolean
+  readonly disfluencies?: boolean
+  readonly filter_profanity?: boolean
+  readonly temperature?: number
+  readonly speaker_options?: { readonly min_speakers_expected?: number; readonly max_speakers_expected?: number }
+  readonly language_detection_options?: {
+    readonly expected_languages?: ReadonlyArray<string>
+    readonly fallback_language?: string
+    readonly code_switching?: boolean
+  }
+  readonly speech_models?: ReadonlyArray<"universal-3-5-pro" | "universal-2" | (string & {})>
+} & Record<string, unknown>
+
+export type Request = TranscriptionRequestFor<AssemblyAITranscriptionOptions>
+
+// ---------------------------------------------------------------------------
+// 2. Token and response schemas
+// ---------------------------------------------------------------------------
+
+export const Token = Schema.Struct({ transcriptID: Schema.String })
+export type Token = Schema.Schema.Type<typeof Token>
+
+const Upload = Schema.Struct({ upload_url: Schema.String })
+
+/** Word and utterance times are milliseconds. */
+const Transcript = Schema.Struct({
+  id: Schema.String,
+  status: Schema.String,
+  text: optionalNull(Schema.String),
+  words: optionalNull(
+    Schema.Array(
+      Schema.Struct({
+        text: Schema.String,
+        start: Schema.Number,
+        end: Schema.Number,
+        confidence: optionalNull(Schema.Number),
+        speaker: optionalNull(Schema.String),
+      }),
+    ),
+  ),
+  utterances: optionalNull(
+    Schema.Array(
+      Schema.Struct({
+        text: Schema.String,
+        start: Schema.Number,
+        end: Schema.Number,
+        speaker: optionalNull(Schema.String),
+      }),
+    ),
+  ),
+  language_code: optionalNull(Schema.String),
+  audio_duration: optionalNull(Schema.Number),
+  speech_model_used: optionalNull(Schema.String),
+  error: optionalNull(Schema.String),
+})
+
+const STATUS = {
+  queued: "queued",
+  processing: "running",
+  completed: "completed",
+  error: "failed",
+} as const satisfies Record<string, Status>
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+const decodeUpload = MediaProtocol.decodeJson(ADAPTER, NAME, Upload)
+
+/** `/v2/transcript` only takes a URL, so inline audio is uploaded to `/v2/upload` first. */
+const prepare = Effect.fn("AssemblyAITranscription.prepare")(function* (request: Request, send: MediaProtocol.Send) {
+  if (request.audio.source.type !== "bytes" && request.audio.source.type !== "base64") return request
+  const audio = yield* MediaInput.inlineBytes(ADAPTER, request.audio)
+  const uploaded = yield* send(UPLOAD_PATH, MediaProtocol.binary(audio, "application/octet-stream")).pipe(
+    Effect.flatMap(decodeUpload),
+  )
+  return { ...request, audio: Media.url(uploaded.value.upload_url, { mediaType: request.audio.mediaType }) }
+})
+
+const fromRequest = Effect.fn("AssemblyAITranscription.fromRequest")(function* (request: Request) {
+  const audio = yield* ProviderShared.mediaReference(request.audio, PROVIDER, NAME)
+  return MediaProtocol.json(
+    mergeJsonRecords(
+      {
+        audio_url: audio.value,
+        speech_models: [request.model.id],
+        language_code: request.language,
+        language_detection: request.language === undefined ? true : undefined,
+        prompt: request.prompt,
+        // Turn-level `utterances`, the only segments AssemblyAI returns, require speaker labels.
+        speaker_labels: request.diarize === true || request.timestamps === "segment" ? true : undefined,
+        speakers_expected: request.speakers,
+      },
+      request.providerOptions,
+      request.http?.body,
+    ) ?? {},
+  )
+})
+
+// ---------------------------------------------------------------------------
+// 6. Response decoding
+// ---------------------------------------------------------------------------
+
+const decodeTranscript = MediaProtocol.decodeJson(ADAPTER, NAME, Transcript)
+
+const decodeStart = Effect.fn("AssemblyAITranscription.decodeStart")(function* (
+  response: HttpClientResponse.HttpClientResponse,
+) {
+  const output = yield* decodeTranscript(response)
+  const status = yield* MediaProtocol.status(STATUS, output.value.status, output)
+  return { token: { transcriptID: output.value.id }, snapshot: { id: output.value.id, status } }
+})
+
+const decodeStatus = Effect.fn("AssemblyAITranscription.decodeStatus")(function* (
+  response: HttpClientResponse.HttpClientResponse,
+  context: MediaProtocol.PollContext<Token>,
+) {
+  const output = yield* decodeTranscript(response)
+  const status = yield* MediaProtocol.status(STATUS, output.value.status, output)
+  return { id: context.token.transcriptID, status }
+})
+
+const seconds = (milliseconds: number) => milliseconds / 1000
+
+const decodeResult = Effect.fn("AssemblyAITranscription.decodeResult")(function* (
+  response: HttpClientResponse.HttpClientResponse,
+  context: MediaProtocol.PollContext<Token>,
+) {
+  const output = yield* decodeTranscript(response)
+  const transcript = output.value
+  const status = yield* MediaProtocol.status(STATUS, transcript.status, output)
+  const error = transcript.error ?? undefined
+  if (status === "failed")
+    return yield* output.ended("failed", `${NAME} transcription failed${error === undefined ? "" : `: ${error}`}`)
+  if (status !== "completed")
+    return yield* output.invalid(`${NAME} transcript ${context.token.transcriptID} has not finished`)
+  const duration = transcript.audio_duration ?? undefined
+  return new TranscriptionResponse({
+    text: transcript.text ?? "",
+    segments: transcript.utterances?.map((utterance) => ({
+      text: utterance.text,
+      startSeconds: seconds(utterance.start),
+      endSeconds: seconds(utterance.end),
+      speaker: utterance.speaker ?? undefined,
+    })),
+    words: transcript.words?.map((word) => ({
+      text: word.text,
+      startSeconds: seconds(word.start),
+      endSeconds: seconds(word.end),
+      speaker: word.speaker ?? undefined,
+      confidence: word.confidence ?? undefined,
+    })),
+    language: transcript.language_code?.toLowerCase(),
+    durationSeconds: duration,
+    usage: duration === undefined ? undefined : { type: "seconds", seconds: duration },
+    providerMetadata: {
+      assemblyai: { transcriptId: transcript.id, speechModel: transcript.speech_model_used ?? undefined },
+    },
+  })
+})
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+const transcriptPath = (token: Token) => `${PATH}/${token.transcriptID}`
+
+export const protocol = MediaProtocol.queued<Request, TranscriptionResponse, Token>({
+  id: ADAPTER,
+  name: NAME,
+  token: Token,
+  start: { prepare, body: { from: fromRequest }, decode: decodeStart },
+  status: { path: transcriptPath, decode: decodeStatus },
+  result: { path: transcriptPath, decode: decodeResult },
+})
+
+export const model = (input: MediaRoute.ModelInput) =>
+  TranscriptionModel.fromRoute<AssemblyAITranscriptionOptions, Token>(
+    { id: ADAPTER, provider: PROVIDER, protocol, baseURL: DEFAULT_BASE_URL, path: PATH },
+    input,
+  )
+
+export const AssemblyAITranscription = {
+  protocol,
+  model,
+} as const
@@ -28,6 +28,7 @@ import { Lifecycle } from "./utils/lifecycle.js"
 import { MistralToolID } from "./utils/mistral-tool-id.js"
 import { ToolSchemaProjection } from "./utils/tool-schema.js"
 import { ToolStream } from "./utils/tool-stream.js"
+import { concatBytes } from "../utils/bytes.js"

 const ADAPTER = "bedrock-converse"

@@ -524,14 +525,7 @@ interface ParserState {
  readonly reasoningRedactedContent: Readonly<Record<number, ReadonlyArray<Uint8Array>>>
 }

-const encodeRedactedContent = (chunks: ReadonlyArray<Uint8Array>) => {
-  const bytes = new Uint8Array(chunks.reduce((total, chunk) => total + chunk.length, 0))
-  chunks.reduce((offset, chunk) => {
-    bytes.set(chunk, offset)
-    return offset + chunk.length
-  }, 0)
-  return Encoding.encodeBase64(bytes)
-}
+const encodeRedactedContent = (chunks: ReadonlyArray<Uint8Array>) => Encoding.encodeBase64(concatBytes(chunks))

 const step = (state: ParserState, event: BedrockEvent) =>
  Effect.gen(function* () {
@@ -0,0 +1,195 @@
+import { Effect, Schema } from "effect"
+import { classifyProviderFailure } from "../provider-error.js"
+import { Framing } from "../route/framing.js"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { AIError, ProviderID, mergeJsonRecords } from "../schema/index.js"
+import { SpeechModel, type SpeechEvent, type SpeechRequestFor } from "../speech.js"
+import { ProviderShared, optionalNull } from "./shared.js"
+import { SpeechStream } from "./utils/speech-stream.js"
+
+const ADAPTER = "cartesia-speech"
+const NAME = "Cartesia"
+const PROVIDER = ProviderID.make("cartesia")
+export const DEFAULT_BASE_URL = "https://api.cartesia.ai"
+export const API_VERSION = "2026-08-14"
+export const BYTES_PATH = "/tts/bytes"
+export const SSE_PATH = "/tts/sse"
+const DEFAULT_SAMPLE_RATE = 44100
+const DEFAULT_BIT_RATE = 128000
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+export type CartesiaSpeechString<Known extends string> = Known | (string & {})
+
+export type CartesiaEncoding = SpeechStream.PcmEncoding
+
+export type CartesiaSpeechOptions = {
+  readonly sampleRate?: 8000 | 16000 | 22050 | 24000 | 44100 | 48000
+  readonly bitRate?: 32000 | 64000 | 96000 | 128000 | 192000
+  readonly encoding?: CartesiaEncoding
+  readonly generation_config?: {
+    readonly volume?: number
+    readonly emotion?: CartesiaSpeechString<"neutral" | "calm" | "angry" | "content" | "sad" | "scared">
+  }
+  readonly pronunciation_dict_id?: string
+} & Record<string, unknown>
+
+export type Request = SpeechRequestFor<CartesiaSpeechOptions>
+
+// ---------------------------------------------------------------------------
+// 3. Streaming event schema
+// ---------------------------------------------------------------------------
+
+/** `phoneme_timestamps` and future record types are ignored. */
+const SseEvent = Schema.Struct({
+  type: Schema.String,
+  data: Schema.optional(Schema.Uint8ArrayFromBase64),
+  word_timestamps: Schema.optional(
+    Schema.Struct({
+      words: Schema.Array(Schema.String),
+      start: Schema.Array(Schema.Number),
+      end: Schema.Array(Schema.Number),
+    }),
+  ),
+  status_code: Schema.optional(Schema.Number),
+  title: Schema.optional(Schema.String),
+  message: Schema.optional(Schema.String),
+  error_code: optionalNull(Schema.String),
+})
+
+const decodeEvent = MediaProtocol.decodeFrame(ADAPTER, NAME, SseEvent)
+
+// ---------------------------------------------------------------------------
+// 4. Parser state
+// ---------------------------------------------------------------------------
+
+interface State extends SpeechStream.Audio {
+  readonly done: boolean
+}
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+/** Timestamps exist only on the SSE endpoint, so a `generate` that asks for them collects an SSE stream. */
+const usesSse = (request: MediaProtocol.Addressed<Request>) => request.mode === "stream" || request.timestamps === true
+
+const CONTAINERS: Readonly<Record<string, "raw" | "wav" | "mp3">> = { pcm: "raw", wav: "wav", mp3: "mp3" }
+
+const outputFormat = Effect.fn("CartesiaSpeech.outputFormat")(function* (request: MediaProtocol.Addressed<Request>) {
+  const sse = usesSse(request)
+  const format = request.format ?? (sse ? "pcm" : "mp3")
+  const container = CONTAINERS[format]
+  if (container === undefined)
+    return yield* SpeechStream.unsupportedFormat(
+      PROVIDER,
+      ADAPTER,
+      `${NAME} supports the pcm, wav, and mp3 formats, not "${format}"`,
+    )
+  if (sse && container !== "raw")
+    return yield* SpeechStream.unsupportedFormat(
+      PROVIDER,
+      ADAPTER,
+      `${NAME} streams and timestamps only raw PCM; request format "pcm" instead of "${format}"`,
+    )
+  const sampleRate = request.providerOptions?.sampleRate ?? DEFAULT_SAMPLE_RATE
+  if (container === "mp3")
+    return { container, sample_rate: sampleRate, bit_rate: request.providerOptions?.bitRate ?? DEFAULT_BIT_RATE }
+  return { container, encoding: request.providerOptions?.encoding ?? "pcm_s16le", sample_rate: sampleRate }
+})
+
+const fromRequest = Effect.fn("CartesiaSpeech.fromRequest")(function* (request: MediaProtocol.Addressed<Request>) {
+  const voice = SpeechStream.voiceID(request.voice)
+  if (voice === undefined)
+    return yield* ProviderShared.invalidRequest(`${NAME} requires a voice id; pass it as \`voice\``)
+  const { sampleRate: _sampleRate, bitRate: _bitRate, encoding: _encoding, ...native } = request.providerOptions ?? {}
+  return MediaProtocol.json(
+    mergeJsonRecords(
+      {
+        model_id: request.model.id,
+        transcript: request.text,
+        voice,
+        output_format: yield* outputFormat(request),
+        language: request.language,
+        generation_config: request.speed === undefined ? undefined : { speed: request.speed },
+        add_timestamps: request.timestamps === true ? true : undefined,
+      },
+      native,
+      request.http?.body,
+    ) ?? {},
+  )
+})
+
+// ---------------------------------------------------------------------------
+// 6. Stream parsing
+// ---------------------------------------------------------------------------
+
+const onEvent = Effect.fn("CartesiaSpeech.onEvent")(function* (state: State, frame: string) {
+  const event = yield* decodeEvent(frame)
+  if (event.type === "chunk" && event.data !== undefined) return SpeechStream.delta(state, event.data)
+  if (event.type === "timestamps" && event.word_timestamps !== undefined) {
+    const words = event.word_timestamps
+    return [state, SpeechStream.timestamps(words.words, words.start, words.end)] as const
+  }
+  if (event.type === "done") return [{ ...state, done: true }, []] as const
+  if (event.type === "error")
+    return yield* new AIError({
+      reason: classifyProviderFailure({
+        message: `${NAME} stream failed${event.title === undefined ? "" : ` (${event.title})`}: ${event.message ?? "unknown error"}`,
+        status: event.status_code,
+        rawBody: frame,
+      }),
+    })
+  return [state, []] as const
+})
+
+const finish = Effect.fn("CartesiaSpeech.finish")(function* (
+  state: State,
+  context: MediaProtocol.ResponseContext<Request>,
+) {
+  if (usesSse(context.request) && !state.done) return yield* MediaProtocol.incomplete(ADAPTER)
+  const format = yield* outputFormat(context.request)
+  return yield* SpeechStream.finish(
+    ADAPTER,
+    state,
+    format.container === "raw"
+      ? SpeechStream.pcm(format.encoding, format.sample_rate)
+      : SpeechStream.container(format.container, format.sample_rate),
+  )
+})
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+export const protocol = MediaProtocol.stream<Request, SpeechEvent, string | Uint8Array, State>({
+  id: ADAPTER,
+  name: NAME,
+  unsupported: ["instructions"],
+  body: { from: fromRequest },
+  frames: (bytes, context) => (usesSse(context.request) ? Framing.sse.frame(bytes) : bytes),
+  initial: () => ({ chunks: [], done: false }),
+  step: SpeechStream.step(onEvent),
+  finish,
+})
+
+export const model = (input: MediaRoute.ModelInput) =>
+  SpeechModel.fromRoute<CartesiaSpeechOptions, string | Uint8Array, State>(
+    {
+      id: ADAPTER,
+      provider: PROVIDER,
+      protocol,
+      baseURL: DEFAULT_BASE_URL,
+      headers: { "Cartesia-Version": API_VERSION },
+      path: ({ request }) => (usesSse(request) ? SSE_PATH : BYTES_PATH),
+    },
+    input,
+  )
+
+export const CartesiaSpeech = {
+  protocol,
+  model,
+} as const
@@ -0,0 +1,144 @@
+import { Effect } from "effect"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
+import { SpeechModel, type SpeechEvent, type SpeechRequestFor } from "../speech.js"
+import { MediaInput } from "./utils/media-input.js"
+import { SpeechStream } from "./utils/speech-stream.js"
+
+const ADAPTER = "deepgram-speech"
+const NAME = "Deepgram"
+const PROVIDER = ProviderID.make("deepgram")
+export const DEFAULT_BASE_URL = "https://api.deepgram.com"
+export const PATH = "/v1/speak"
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+export type DeepgramSpeechString<Known extends string> = Known | (string & {})
+
+export type DeepgramEncoding = DeepgramSpeechString<"linear16" | "mulaw" | "alaw" | "mp3" | "opus" | "flac" | "aac">
+
+export type DeepgramSpeechOptions = {
+  readonly encoding?: DeepgramEncoding
+  readonly container?: DeepgramSpeechString<"wav" | "ogg" | "none">
+  readonly sampleRate?: number
+  readonly bitRate?: number
+  readonly mip_opt_out?: boolean
+  readonly tag?: string
+} & Record<string, unknown>
+
+export type Request = SpeechRequestFor<DeepgramSpeechOptions>
+
+// ---------------------------------------------------------------------------
+// 4. Parser state
+// ---------------------------------------------------------------------------
+
+type State = SpeechStream.Audio
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+const FORMATS: Readonly<Record<string, { readonly encoding: string; readonly container?: string }>> = {
+  mp3: { encoding: "mp3" },
+  wav: { encoding: "linear16", container: "wav" },
+  pcm: { encoding: "linear16", container: "none" },
+  opus: { encoding: "opus" },
+  flac: { encoding: "flac" },
+  aac: { encoding: "aac" },
+}
+
+const audioFormat = (request: Request) => {
+  const format = request.format === undefined ? undefined : FORMATS[request.format]
+  return {
+    encoding: request.providerOptions?.encoding ?? format?.encoding,
+    container: request.providerOptions?.container ?? format?.container,
+  }
+}
+
+const queryParameters = (request: Request) => {
+  const { encoding: _encoding, container: _container, sampleRate, bitRate, ...native } = request.providerOptions ?? {}
+  return MediaInput.query(ADAPTER, {
+    ...native,
+    model: request.model.id,
+    ...audioFormat(request),
+    sample_rate: sampleRate,
+    bit_rate: bitRate,
+    speed: request.speed,
+  })
+}
+
+const fromRequest = Effect.fn("DeepgramSpeech.fromRequest")(function* (request: Request) {
+  if (
+    request.format !== undefined &&
+    FORMATS[request.format] === undefined &&
+    request.providerOptions?.encoding === undefined
+  )
+    return yield* SpeechStream.unsupportedFormat(
+      PROVIDER,
+      ADAPTER,
+      `${NAME} has no encoding for format "${request.format}"; pass providerOptions.encoding`,
+    )
+  return MediaProtocol.json(
+    mergeJsonRecords({ text: request.text }, request.http?.body) ?? {},
+    yield* queryParameters(request),
+  )
+})
+
+// ---------------------------------------------------------------------------
+// 6. Stream parsing
+// ---------------------------------------------------------------------------
+
+const HEADERLESS_ENCODINGS: Readonly<Record<string, SpeechStream.PcmEncoding>> = {
+  linear16: "pcm_s16le",
+  mulaw: "pcm_mulaw",
+  alaw: "pcm_alaw",
+}
+
+const finish = (state: State, context: MediaProtocol.ResponseContext<Request>) => {
+  const headers = context.http.headers
+  const mediaType = headers["content-type"]
+  const format = audioFormat(context.request)
+  const encoding = HEADERLESS_ENCODINGS[format.encoding ?? ""]
+  const requestID = headers["dg-request-id"]
+  const modelName = headers["dg-model-name"]
+  return SpeechStream.finish(ADAPTER, state, {
+    ...(format.container === "none" && encoding !== undefined
+      ? SpeechStream.pcm(encoding, SpeechStream.sampleRate(mediaType), mediaType)
+      : // Deepgram's default encoding is MP3; WAV is a container around any encoding.
+        { mediaType, info: { format: format.container === "wav" ? "wav" : (format.encoding ?? "mp3") } }),
+    usage: SpeechStream.headerUsage("characters", headers["dg-char-count"]),
+    providerMetadata:
+      requestID === undefined && modelName === undefined
+        ? undefined
+        : { deepgram: { requestId: requestID, modelName } },
+  })
+}
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+export const protocol = MediaProtocol.stream<Request, SpeechEvent, Uint8Array, State>({
+  id: ADAPTER,
+  name: NAME,
+  unsupported: ["voice", "language", "instructions", "timestamps"],
+  body: { from: fromRequest },
+  frames: (bytes) => bytes,
+  initial: () => ({ chunks: [] }),
+  step: (state, frame) => Effect.succeed(SpeechStream.delta(state, frame)),
+  finish,
+})
+
+export const model = (input: MediaRoute.ModelInput) =>
+  SpeechModel.fromRoute<DeepgramSpeechOptions, Uint8Array, State>(
+    { id: ADAPTER, provider: PROVIDER, protocol, baseURL: DEFAULT_BASE_URL, path: PATH },
+    input,
+  )
+
+export const DeepgramSpeech = {
+  protocol,
+  model,
+} as const
@@ -0,0 +1,191 @@
+import { Effect, Schema } from "effect"
+import type { HttpClientResponse } from "effect/unstable/http"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
+import { TranscriptionModel, TranscriptionResponse, type TranscriptionRequestFor } from "../transcription.js"
+import { ProviderShared } from "./shared.js"
+import { MediaInput } from "./utils/media-input.js"
+
+const ADAPTER = "deepgram-transcription"
+const NAME = "Deepgram"
+const PROVIDER = ProviderID.make("deepgram")
+export const DEFAULT_BASE_URL = "https://api.deepgram.com"
+export const PATH = "/v1/listen"
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+export type DeepgramTranscriptionOptions = {
+  readonly smart_format?: boolean
+  readonly punctuate?: boolean
+  readonly paragraphs?: boolean
+  readonly utterances?: boolean
+  readonly detect_language?: boolean | ReadonlyArray<string>
+  readonly keyterm?: ReadonlyArray<string>
+  readonly diarize_model?: "latest" | "v1" | "v2" | (string & {})
+  readonly filler_words?: boolean
+  readonly numerals?: boolean
+  readonly mip_opt_out?: boolean
+  readonly tag?: string | ReadonlyArray<string>
+} & Record<string, unknown>
+
+export type Request = TranscriptionRequestFor<DeepgramTranscriptionOptions>
+
+// ---------------------------------------------------------------------------
+// 2. Response schema
+// ---------------------------------------------------------------------------
+
+const Word = Schema.Struct({
+  word: Schema.String,
+  start: Schema.Number,
+  end: Schema.Number,
+  confidence: Schema.optional(Schema.Number),
+  speaker: Schema.optional(Schema.Number),
+  punctuated_word: Schema.optional(Schema.String),
+})
+
+const ListenResponse = Schema.Struct({
+  metadata: Schema.optional(
+    Schema.Struct({ request_id: Schema.optional(Schema.String), duration: Schema.optional(Schema.Number) }),
+  ),
+  results: Schema.Struct({
+    channels: Schema.Array(
+      Schema.Struct({
+        alternatives: Schema.optional(
+          Schema.Array(Schema.Struct({ transcript: Schema.String, words: Schema.optional(Schema.Array(Word)) })),
+        ),
+        detected_language: Schema.optional(Schema.String),
+      }),
+    ),
+    utterances: Schema.optional(
+      Schema.Array(
+        Schema.Struct({
+          start: Schema.Number,
+          end: Schema.Number,
+          transcript: Schema.String,
+          speaker: Schema.optional(Schema.Number),
+          words: Schema.optional(Schema.Array(Word)),
+        }),
+      ),
+    ),
+  }),
+})
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+const query = (request: Request) =>
+  MediaInput.query(
+    ADAPTER,
+    mergeJsonRecords(
+      {
+        model: request.model.id,
+        smart_format: true,
+        language: request.language,
+        // Deepgram assumes English unless asked to detect, unlike the other routes' auto-detection.
+        detect_language: request.language === undefined ? true : undefined,
+        // `diarize=true` is deprecated in favor of choosing a diarization model.
+        diarize_model: request.diarize === true ? "latest" : undefined,
+        utterances: request.diarize === true || request.timestamps === "segment" ? true : undefined,
+      },
+      request.providerOptions,
+    ) ?? {},
+  )
+
+const fromRequest = Effect.fn("DeepgramTranscription.fromRequest")(function* (request: Request) {
+  const url = ProviderShared.mediaUrl(request.audio)
+  if (url !== undefined)
+    return MediaProtocol.json(mergeJsonRecords({ url }, request.http?.body) ?? {}, yield* query(request))
+  if (request.http?.body !== undefined)
+    return yield* ProviderShared.invalidRequest(`${NAME} sends inline audio as the raw body, so http.body cannot apply`)
+  const audio = yield* MediaInput.inlineBytes(ADAPTER, request.audio)
+  return MediaProtocol.binary(audio, request.audio.mediaType, yield* query(request))
+})
+
+// ---------------------------------------------------------------------------
+// 6. Response decoding
+// ---------------------------------------------------------------------------
+
+const decodeListen = MediaProtocol.decodeJson(ADAPTER, NAME, ListenResponse)
+
+const speaker = (value: number | undefined) => (value === undefined ? undefined : String(value))
+
+const wordText = (word: typeof Word.Type) => word.punctuated_word ?? word.word
+
+// Utterances split on pauses, not speakers: the v2 diarizer labels a whole utterance with one speaker even when its
+// words change speaker, so segments split each utterance at speaker changes.
+const speakerTurns = (words: ReadonlyArray<typeof Word.Type>) =>
+  words.reduce<Array<Array<typeof Word.Type>>>((turns, word) => {
+    const last = turns.at(-1)
+    if (last === undefined || last[0].speaker !== word.speaker) return [...turns, [word]]
+    last.push(word)
+    return turns
+  }, [])
+
+const decodeResponse = Effect.fn("DeepgramTranscription.decodeResponse")(function* (
+  response: HttpClientResponse.HttpClientResponse,
+) {
+  const output = yield* decodeListen(response)
+  const channel = output.value.results.channels[0]
+  const alternative = channel?.alternatives?.[0]
+  if (alternative === undefined) return yield* output.invalid(`${NAME} returned no transcript`)
+  const duration = output.value.metadata?.duration
+  const requestID = output.value.metadata?.request_id
+  return new TranscriptionResponse({
+    text: alternative.transcript,
+    segments: output.value.results.utterances?.flatMap((utterance) =>
+      utterance.words === undefined || utterance.words.length === 0
+        ? [
+            {
+              text: utterance.transcript,
+              startSeconds: utterance.start,
+              endSeconds: utterance.end,
+              speaker: speaker(utterance.speaker),
+            },
+          ]
+        : speakerTurns(utterance.words).map((turn) => ({
+            text: turn.map(wordText).join(" "),
+            startSeconds: turn[0].start,
+            endSeconds: turn[turn.length - 1].end,
+            speaker: speaker(turn[0].speaker),
+          })),
+    ),
+    words: alternative.words?.map((word) => ({
+      text: wordText(word),
+      startSeconds: word.start,
+      endSeconds: word.end,
+      speaker: speaker(word.speaker),
+      confidence: word.confidence,
+    })),
+    language: channel?.detected_language?.toLowerCase(),
+    durationSeconds: duration,
+    usage: duration === undefined ? undefined : { type: "seconds", seconds: duration },
+    providerMetadata: requestID === undefined ? undefined : { deepgram: { requestId: requestID } },
+  })
+})
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+export const protocol = MediaProtocol.inline<Request, TranscriptionResponse>({
+  id: ADAPTER,
+  name: NAME,
+  unsupported: ["prompt", "speakers"],
+  body: { from: fromRequest },
+  response: { decode: decodeResponse },
+})
+
+export const model = (input: MediaRoute.ModelInput) =>
+  TranscriptionModel.fromRoute<DeepgramTranscriptionOptions>(
+    { id: ADAPTER, provider: PROVIDER, protocol, baseURL: DEFAULT_BASE_URL, path: PATH },
+    input,
+  )
+
+export const DeepgramTranscription = {
+  protocol,
+  model,
+} as const
@@ -0,0 +1,218 @@
+import { Effect, Schema } from "effect"
+import { Framing } from "../route/framing.js"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
+import { SpeechModel, type SpeechEvent, type SpeechRequestFor } from "../speech.js"
+import { ProviderShared, optionalNull } from "./shared.js"
+import { SpeechStream } from "./utils/speech-stream.js"
+
+const ADAPTER = "elevenlabs-speech"
+const NAME = "ElevenLabs"
+const PROVIDER = ProviderID.make("elevenlabs")
+export const DEFAULT_BASE_URL = "https://api.elevenlabs.io"
+export const PATH = "/v1/text-to-speech"
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+export type ElevenLabsSpeechString<Known extends string> = Known | (string & {})
+
+export type ElevenLabsOutputFormat = ElevenLabsSpeechString<
+  | "mp3_22050_32"
+  | "mp3_24000_48"
+  | "mp3_44100_32"
+  | "mp3_44100_64"
+  | "mp3_44100_96"
+  | "mp3_44100_128"
+  | "mp3_44100_192"
+  | "pcm_8000"
+  | "pcm_16000"
+  | "pcm_22050"
+  | "pcm_24000"
+  | "pcm_32000"
+  | "pcm_44100"
+  | "pcm_48000"
+  | "wav_8000"
+  | "wav_16000"
+  | "wav_22050"
+  | "wav_24000"
+  | "wav_32000"
+  | "wav_44100"
+  | "wav_48000"
+  | "ulaw_8000"
+  | "alaw_8000"
+  | "opus_48000_32"
+  | "opus_48000_64"
+  | "opus_48000_96"
+  | "opus_48000_128"
+  | "opus_48000_192"
+>
+
+export type ElevenLabsSpeechOptions = {
+  readonly outputFormat?: ElevenLabsOutputFormat
+  readonly voice_settings?: {
+    readonly stability?: number
+    readonly similarity_boost?: number
+    readonly style?: number
+    readonly use_speaker_boost?: boolean
+  }
+  readonly seed?: number
+  readonly apply_text_normalization?: ElevenLabsSpeechString<"auto" | "on" | "off">
+} & Record<string, unknown>
+
+export type Request = SpeechRequestFor<ElevenLabsSpeechOptions>
+
+// ---------------------------------------------------------------------------
+// 3. Streaming event schema
+// ---------------------------------------------------------------------------
+
+const Alignment = Schema.Struct({
+  characters: Schema.Array(Schema.String),
+  character_start_times_seconds: Schema.Array(Schema.Number),
+  character_end_times_seconds: Schema.Array(Schema.Number),
+})
+
+const TimestampedAudio = Schema.Struct({
+  audio_base64: Schema.Uint8ArrayFromBase64,
+  alignment: optionalNull(Alignment),
+})
+
+const decodeRecord = MediaProtocol.decodeFrame(ADAPTER, NAME, TimestampedAudio)
+
+// ---------------------------------------------------------------------------
+// 4. Parser state
+// ---------------------------------------------------------------------------
+
+type State = SpeechStream.Audio
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+const OUTPUT_FORMATS: Readonly<Record<string, string>> = {
+  mp3: "mp3_44100_128",
+  pcm: "pcm_24000",
+  wav: "wav_24000",
+  opus: "opus_48000_64",
+}
+
+/** WAV is served only by the non-streaming endpoints. */
+const outputFormat = Effect.fn("ElevenLabsSpeech.outputFormat")(function* (request: MediaProtocol.Addressed<Request>) {
+  const format = request.providerOptions?.outputFormat ?? OUTPUT_FORMATS[request.format ?? "mp3"]
+  if (format === undefined)
+    return yield* SpeechStream.unsupportedFormat(
+      PROVIDER,
+      ADAPTER,
+      `${NAME} has no default output format for "${request.format}"; pass providerOptions.outputFormat`,
+    )
+  if (request.mode === "stream" && format.startsWith("wav_"))
+    return yield* SpeechStream.unsupportedFormat(
+      PROVIDER,
+      ADAPTER,
+      `${NAME} streams mp3, pcm, opus, ulaw, and alaw but not "${format}"; use generate for WAV`,
+    )
+  return format
+})
+
+const fromRequest = Effect.fn("ElevenLabsSpeech.fromRequest")(function* (request: MediaProtocol.Addressed<Request>) {
+  if (request.voice === undefined)
+    return yield* ProviderShared.invalidRequest(`${NAME} requires a voice id; pass it as \`voice\``)
+  const { outputFormat: _outputFormat, ...native } = request.providerOptions ?? {}
+  return MediaProtocol.json(
+    mergeJsonRecords(
+      {
+        text: request.text,
+        model_id: request.model.id,
+        language_code: request.language,
+        voice_settings: request.speed === undefined ? undefined : { speed: request.speed },
+      },
+      native,
+      request.http?.body,
+    ) ?? {},
+    { output_format: yield* outputFormat(request) },
+  )
+})
+
+const path = (request: MediaProtocol.Addressed<Request>) =>
+  `${PATH}/${encodeURIComponent(SpeechStream.voiceID(request.voice) ?? "")}${request.mode === "stream" ? "/stream" : ""}${
+    request.timestamps === true ? "/with-timestamps" : ""
+  }`
+
+// ---------------------------------------------------------------------------
+// 6. Stream parsing
+// ---------------------------------------------------------------------------
+
+const onRecord = Effect.fn("ElevenLabsSpeech.onRecord")(function* (state: State, frame: string) {
+  const record = yield* decodeRecord(frame)
+  const [next, events] = SpeechStream.delta(state, record.audio_base64)
+  const alignment = record.alignment
+  if (!alignment) return [next, events] as const
+  return [
+    next,
+    [
+      ...events,
+      ...SpeechStream.timestamps(
+        alignment.characters,
+        alignment.character_start_times_seconds,
+        alignment.character_end_times_seconds,
+      ),
+    ],
+  ] as const
+})
+
+const PCM_CODECS: Readonly<Record<string, SpeechStream.PcmEncoding>> = {
+  pcm: "pcm_s16le",
+  ulaw: "pcm_mulaw",
+  alaw: "pcm_alaw",
+}
+
+const describeOutput = (format: string) => {
+  const [codec = format, rate] = format.split("_")
+  const sampleRate = rate === undefined ? undefined : Number(rate)
+  const encoding = PCM_CODECS[codec]
+  return encoding === undefined ? SpeechStream.container(codec, sampleRate) : SpeechStream.pcm(encoding, sampleRate)
+}
+
+const finish = Effect.fn("ElevenLabsSpeech.finish")(function* (
+  state: State,
+  context: MediaProtocol.ResponseContext<Request>,
+) {
+  const requestID = context.http.headers["request-id"]
+  return yield* SpeechStream.finish(ADAPTER, state, {
+    ...describeOutput(yield* outputFormat(context.request)),
+    // `character-cost` is billed credits, not a character count (3 for 20 characters on `eleven_flash_v2_5`).
+    usage: SpeechStream.headerUsage("credits", context.http.headers["character-cost"]),
+    providerMetadata: requestID === undefined ? undefined : { elevenlabs: { requestId: requestID } },
+  })
+})
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+export const protocol = MediaProtocol.stream<Request, SpeechEvent, string | Uint8Array, State>({
+  id: ADAPTER,
+  name: NAME,
+  unsupported: ["instructions"],
+  body: { from: fromRequest },
+  frames: (bytes, context) => {
+    if (context.request.timestamps !== true) return bytes
+    return context.request.mode === "stream" ? Framing.lines.frame(bytes) : Framing.document.frame(bytes)
+  },
+  initial: () => ({ chunks: [] }),
+  step: SpeechStream.step(onRecord),
+  finish,
+})
+
+export const model = (input: MediaRoute.ModelInput) =>
+  SpeechModel.fromRoute<ElevenLabsSpeechOptions, string | Uint8Array, State>(
+    { id: ADAPTER, provider: PROVIDER, protocol, baseURL: DEFAULT_BASE_URL, path: ({ request }) => path(request) },
+    input,
+  )
+
+export const ElevenLabsSpeech = {
+  protocol,
+  model,
+} as const
@@ -0,0 +1,198 @@
+import { Effect, Schema } from "effect"
+import type { HttpClientResponse } from "effect/unstable/http"
+import type { Status } from "../generation.js"
+import { Media } from "../media.js"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
+import { VideoModel, VideoResponse, type VideoRequestFor } from "../video.js"
+import { ProviderShared, optionalNull } from "./shared.js"
+
+const ADAPTER = "fal-video"
+const NAME = "fal Video"
+const PROVIDER = ProviderID.make("fal")
+export const DEFAULT_BASE_URL = "https://queue.fal.run"
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+export type FalVideoString<Known extends string> = Known | (string & {})
+
+/**
+ * Provider-native input. fal video endpoints are model-specific: `duration` is a string enum whose values differ per
+ * model (`"8s"` for Veo, `"5"` for Kling), and last-frame fields are named per model (`end_image_url`,
+ * `last_frame_url`, `tail_image_url`), so those pass through here instead of lowering from common fields.
+ */
+export type FalVideoOptions = {
+  readonly duration?: FalVideoString<"4s" | "6s" | "8s" | "5" | "10">
+} & Record<string, unknown>
+
+export type Request = VideoRequestFor<FalVideoOptions>
+
+// ---------------------------------------------------------------------------
+// 2. Token and response schemas
+// ---------------------------------------------------------------------------
+
+/** fal hands back absolute follow-up URLs on submit; they are authoritative for status, result, and cancel. */
+export const Token = Schema.Struct({
+  requestID: Schema.String,
+  statusURL: Schema.String,
+  responseURL: Schema.String,
+  cancelURL: Schema.String,
+})
+export type Token = Schema.Schema.Type<typeof Token>
+
+const StartResponse = Schema.Struct({
+  request_id: Schema.String,
+  status_url: Schema.String,
+  response_url: Schema.String,
+  cancel_url: Schema.String,
+  queue_position: optionalNull(Schema.Number),
+})
+
+const QueueStatus = Schema.Struct({
+  status: Schema.String,
+  queue_position: optionalNull(Schema.Number),
+  error: optionalNull(Schema.Unknown),
+})
+
+const QueueResult = Schema.StructWithRest(
+  Schema.Struct({
+    video: Schema.Struct({
+      url: Schema.String,
+      content_type: optionalNull(Schema.String),
+      file_name: optionalNull(Schema.String),
+      file_size: optionalNull(Schema.Number),
+    }),
+    seed: optionalNull(Schema.Number),
+  }),
+  [Schema.Record(Schema.String, Schema.Unknown)],
+)
+
+const STATUS = {
+  IN_QUEUE: "queued",
+  IN_PROGRESS: "running",
+  COMPLETED: "completed",
+} as const satisfies Record<string, Status>
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+// fal accepts public URLs and data URIs; there is no provider file handle to forward.
+const mediaUrl = (asset: Media.Asset) =>
+  ProviderShared.mediaReference(asset, undefined, NAME).pipe(Effect.map((reference) => reference.value))
+
+const fromRequest = Effect.fn("FalVideo.fromRequest")(function* (request: Request) {
+  if (request.frames?.last !== undefined)
+    return yield* ProviderShared.unsupportedOperation({
+      operation: "video.frames.last",
+      provider: PROVIDER,
+      route: ADAPTER,
+      message: `${NAME} names the last frame per model; pass it through providerOptions (e.g. end_image_url) instead of frames.last`,
+    })
+  const imageUrl = request.frames?.first === undefined ? undefined : yield* mediaUrl(request.frames.first)
+  const videoUrl = request.video === undefined ? undefined : yield* mediaUrl(request.video)
+  return MediaProtocol.json(
+    mergeJsonRecords(
+      {
+        prompt: request.prompt,
+        negative_prompt: request.negativePrompt,
+        seed: request.seed,
+        aspect_ratio: request.aspectRatio,
+        resolution: request.resolution,
+        generate_audio: request.audio,
+        image_url: imageUrl,
+        video_url: videoUrl,
+      },
+      request.providerOptions,
+      request.http?.body,
+    ) ?? {},
+  )
+})
+
+// ---------------------------------------------------------------------------
+// 6. Response decoding
+// ---------------------------------------------------------------------------
+
+const decodeStart = MediaProtocol.decodeStarted(ADAPTER, NAME, StartResponse, (value) => ({
+  token: {
+    requestID: value.request_id,
+    statusURL: value.status_url,
+    responseURL: value.response_url,
+    cancelURL: value.cancel_url,
+  },
+  snapshot: { id: value.request_id, status: "queued", position: value.queue_position ?? undefined },
+}))
+
+const decodeQueueStatus = MediaProtocol.decodeJson(ADAPTER, NAME, QueueStatus)
+const decodeQueueResult = MediaProtocol.decodeJson(ADAPTER, NAME, QueueResult)
+
+const decodeStatus = Effect.fn("FalVideo.decodeStatus")(function* (
+  response: HttpClientResponse.HttpClientResponse,
+  context: MediaProtocol.PollContext<Token>,
+) {
+  const output = yield* decodeQueueStatus(response)
+  const decoded = output.value
+  const status = yield* MediaProtocol.status(STATUS, decoded.status, output)
+  // fal reports request failures as COMPLETED with an `error`; the response endpoint carries the details.
+  const failed = status === "completed" && decoded.error !== undefined && decoded.error !== null
+  return {
+    id: context.token.requestID,
+    status: failed ? "failed" : status,
+    position: status === "queued" ? (decoded.queue_position ?? undefined) : undefined,
+  }
+})
+
+const decodeResult = Effect.fn("FalVideo.decodeResult")(function* (
+  response: HttpClientResponse.HttpClientResponse,
+  context: MediaProtocol.PollContext<Token>,
+) {
+  const output = yield* decodeQueueResult(response)
+  const { video, seed, ...rest } = output.value
+  return new VideoResponse({
+    videos: [Media.url(video.url, { mediaType: video.content_type ?? "video/mp4" })],
+    providerMetadata: {
+      fal: {
+        requestId: context.token.requestID,
+        seed: seed ?? undefined,
+        fileName: video.file_name ?? undefined,
+        fileSize: video.file_size ?? undefined,
+        ...rest,
+      },
+    },
+  })
+})
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+export const protocol = MediaProtocol.queued<Request, VideoResponse, Token>({
+  id: ADAPTER,
+  name: NAME,
+  token: Token,
+  unsupported: ["n", "durationSeconds", "references"],
+  start: { body: { from: fromRequest }, decode: decodeStart },
+  status: { path: (token) => token.statusURL, decode: decodeStatus },
+  result: { path: (token) => token.responseURL, decode: decodeResult },
+  cancel: { method: "PUT", path: (token) => token.cancelURL },
+})
+
+export const model = (input: MediaRoute.ModelInput) =>
+  VideoModel.fromRoute<FalVideoOptions, Token>(
+    {
+      id: ADAPTER,
+      provider: PROVIDER,
+      protocol,
+      baseURL: DEFAULT_BASE_URL,
+      path: ({ request }) => `/${request.model.id}`,
+    },
+    input,
+  )
+
+export const FalVideo = {
+  protocol,
+  model,
+} as const
@@ -22,6 +22,7 @@ import {
 import { classifyProviderFailure } from "../provider-error.js"
 import { Media } from "../media.js"
 import { JsonObject, knownString, lenient, optionalArray, optionalNull, ProviderShared } from "./shared.js"
+import { GeminiGenerateContent } from "./utils/gemini-generate-content.js"
 import { GeminiToolSchema } from "./utils/gemini-tool-schema.js"
 import { Lifecycle } from "./utils/lifecycle.js"
 import { ToolSchemaProjection } from "./utils/tool-schema.js"
@@ -305,14 +306,9 @@ const lowerToolConfig = (toolChoice: NonNullable<LLMRequest["toolChoice"]>) =>
    tool: (name) => ({ functionCallingConfig: { mode: "ANY" as const, allowedFunctionNames: [name] } }),
  })

-// Gemini does not fetch public URLs; inline payloads and Gemini Files references are the accepted inputs.
 const lowerContentPart = Effect.fn("Gemini.lowerContentPart")(function* (part: TextPart | MediaPart) {
  if (part.type === "text") return { text: part.text }
-  const source = part.media.source
-  if (source.type === "ref" && source.provider === "google")
-    return { fileData: { mimeType: part.media.mediaType, fileUri: source.id } }
-  const media = yield* ProviderShared.requireInlineMedia("Gemini", part.media)
-  return { inlineData: { mimeType: media.mime, data: media.base64 } }
+  return yield* GeminiGenerateContent.mediaPart("Gemini", part.media)
 })

 const providerMetadata = (key: string, metadata: Record<string, unknown>): ProviderMetadata => ({ [key]: metadata })
@@ -1,11 +1,11 @@
 import { Effect, Schema } from "effect"
 import type { HttpClientResponse } from "effect/unstable/http"
 import { ImageModel, ImageResponse, type ImageRequestFor } from "../image.js"
-import { Media } from "../media.js"
 import { MediaProtocol } from "../route/media-protocol.js"
 import { MediaRoute } from "../route/media.js"
-import { ProviderID, mergeJsonRecords, type AIError } from "../schema/index.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
 import { ProviderShared } from "./shared.js"
+import { GeminiGenerateContent } from "./utils/gemini-generate-content.js"
 import { MediaInput } from "./utils/media-input.js"

 const ADAPTER = "google-images"
@@ -102,25 +102,6 @@ const generationConfig = (request: Request) => {
  )
 }

-// Gemini does not fetch public URLs; inline payloads or Gemini Files references are the only accepted inputs.
-const imagePart = (asset: Media.Asset): Effect.Effect<Record<string, unknown>, AIError> => {
-  const inline = asset.inline()
-  if (inline) return Effect.succeed({ inlineData: { mimeType: inline.mime, data: inline.base64 } })
-  const id = MediaInput.refID(asset, PROVIDER)
-  if (id) return Effect.succeed({ fileData: { mimeType: asset.mediaType, fileUri: id } })
-  if (asset.source.type === "ref")
-    return Effect.fail(
-      ProviderShared.invalidRequest(
-        "Google generateContent requires Gemini file references rather than other providers' file IDs",
-      ),
-    )
-  return Effect.fail(
-    ProviderShared.invalidRequest(
-      "Google generateContent does not fetch public image URLs; use bytes, a data URL, or a Gemini file reference",
-    ),
-  )
-}
-
 const fromRequest = Effect.fn("GoogleImages.fromRequest")(function* (request: Request) {
  if (request.n !== undefined && request.n > 1)
    return yield* ProviderShared.unsupportedOperation({
@@ -129,7 +110,7 @@ const fromRequest = Effect.fn("GoogleImages.fromRequest")(function* (request: Re
      route: ADAPTER,
      message: `${NAME} generates one image per request; call it once per image instead of n=${request.n}`,
    })
-  const parts = yield* Effect.forEach(request.images ?? [], imagePart)
+  const parts = yield* Effect.forEach(request.images ?? [], (image) => GeminiGenerateContent.mediaPart(NAME, image))
  return MediaProtocol.json(
    mergeJsonRecords(
      {
@@ -0,0 +1,144 @@
+import { Effect, Schema } from "effect"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
+import { SpeechModel, type SpeechEvent, type SpeechRequestFor } from "../speech.js"
+import { GeminiGenerateContent } from "./utils/gemini-generate-content.js"
+import { SpeechStream } from "./utils/speech-stream.js"
+
+const ADAPTER = "google-speech"
+const NAME = "Google Speech"
+const PROVIDER = ProviderID.make("google")
+export const DEFAULT_BASE_URL = "https://generativelanguage.googleapis.com/v1beta"
+const DEFAULT_SAMPLE_RATE = 24000
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+/** Style is directed in the text itself, and `speechConfig.multiSpeakerVoiceConfig` excludes `voice`. */
+export type GoogleSpeechOptions = {
+  readonly temperature?: number
+  readonly seed?: number
+  readonly speechConfig?: {
+    readonly multiSpeakerVoiceConfig?: {
+      readonly speakerVoiceConfigs: ReadonlyArray<{
+        readonly speaker: string
+        readonly voiceConfig: { readonly prebuiltVoiceConfig: { readonly voiceName: string } }
+      }>
+    }
+  }
+} & Record<string, unknown>
+
+export type Request = SpeechRequestFor<GoogleSpeechOptions>
+
+// ---------------------------------------------------------------------------
+// 3. Streaming event schema
+// ---------------------------------------------------------------------------
+
+const GenerateContentChunk = GeminiGenerateContent.chunk(
+  Schema.Struct({
+    text: Schema.optional(Schema.String),
+    inlineData: Schema.optional(Schema.Struct({ mimeType: Schema.String, data: Schema.Uint8ArrayFromBase64 })),
+  }),
+)
+
+const decodeChunk = MediaProtocol.decodeFrame(ADAPTER, NAME, GenerateContentChunk)
+
+// ---------------------------------------------------------------------------
+// 4. Parser state
+// ---------------------------------------------------------------------------
+
+interface State extends SpeechStream.Audio, GeminiGenerateContent.Metadata {
+  readonly mimeType?: string
+}
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+const fromRequest = Effect.fn("GoogleSpeech.fromRequest")(function* (request: MediaProtocol.Addressed<Request>) {
+  if (request.format !== undefined && request.format !== "pcm")
+    return yield* SpeechStream.unsupportedFormat(
+      PROVIDER,
+      ADAPTER,
+      `${NAME} only returns raw PCM; request format "pcm" or omit it, then wrap the samples yourself`,
+    )
+  const voiceName = SpeechStream.voiceID(request.voice)
+  return MediaProtocol.json(
+    mergeJsonRecords(
+      {
+        contents: [{ role: "user", parts: [{ text: request.text }] }],
+        generationConfig: mergeJsonRecords(
+          {
+            responseModalities: ["AUDIO"],
+            speechConfig: {
+              voiceConfig: voiceName === undefined ? undefined : { prebuiltVoiceConfig: { voiceName } },
+              languageCode: request.language,
+            },
+          },
+          request.providerOptions,
+        ),
+      },
+      request.http?.body,
+    ) ?? {},
+  )
+})
+
+// ---------------------------------------------------------------------------
+// 6. Stream parsing
+// ---------------------------------------------------------------------------
+
+const step = Effect.fn("GoogleSpeech.step")(function* (state: State, frame: string) {
+  const chunk = yield* decodeChunk(frame)
+  const blocked = GeminiGenerateContent.blocked(NAME, chunk, frame)
+  if (blocked !== undefined) return yield* blocked
+  const audio = (chunk.candidates?.[0]?.content?.parts ?? []).flatMap((part) =>
+    part.inlineData === undefined ? [] : [part.inlineData],
+  )
+  const next: State = { ...GeminiGenerateContent.track(state, chunk), mimeType: state.mimeType ?? audio[0]?.mimeType }
+  return [next, audio.flatMap((part) => SpeechStream.delta(next, part.data)[1])] as const
+})
+
+const finish = (state: State) => {
+  const sampleRate = SpeechStream.sampleRate(state.mimeType) ?? DEFAULT_SAMPLE_RATE
+  return SpeechStream.finish(ADAPTER, state, {
+    ...SpeechStream.pcm("pcm_s16le", sampleRate, state.mimeType ?? `audio/L16;codec=pcm;rate=${sampleRate}`),
+    usage: GeminiGenerateContent.usage(state.usage),
+    providerMetadata: GeminiGenerateContent.providerMetadata(state),
+    detail: state.finishReason === undefined ? undefined : `finish reason: ${state.finishReason}`,
+  })
+}
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+export const protocol = MediaProtocol.stream<Request, SpeechEvent, string, State>({
+  id: ADAPTER,
+  name: NAME,
+  unsupported: ["instructions", "speed", "timestamps"],
+  body: { from: fromRequest },
+  frames: (bytes, context) => GeminiGenerateContent.frames(bytes, context.request.mode),
+  initial: () => ({ chunks: [] }),
+  step,
+  finish,
+})
+
+export const model = (input: MediaRoute.ModelInput) =>
+  SpeechModel.fromRoute<GoogleSpeechOptions, string, State>(
+    {
+      id: ADAPTER,
+      provider: PROVIDER,
+      protocol,
+      baseURL: DEFAULT_BASE_URL,
+      // Only `gemini-3.1-flash-tts-preview` and later stream; earlier TTS models reject `streamGenerateContent`.
+      path: ({ request }) => GeminiGenerateContent.path(request.model.id, request.mode),
+    },
+    input,
+  )
+
+export const GoogleSpeech = {
+  protocol,
+  model,
+} as const
@@ -0,0 +1,214 @@
+import { Effect, Schema, SchemaGetter } from "effect"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
+import {
+  TranscriptionFinishEvent,
+  TranscriptionModel,
+  TranscriptionSegmentEvent,
+  TranscriptionTextDeltaEvent,
+  type TranscriptionRequestFor,
+  type TranscriptionSegment,
+  type TranscriptionWord,
+  type TranscriptionEvent,
+} from "../transcription.js"
+import { ProviderShared } from "./shared.js"
+import { GeminiGenerateContent } from "./utils/gemini-generate-content.js"
+
+const ADAPTER = "google-transcription"
+const NAME = "Google Transcription"
+const PROVIDER = ProviderID.make("google")
+export const DEFAULT_BASE_URL = "https://generativelanguage.googleapis.com/v1beta"
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+/**
+ * Merged into `generationConfig`. The API rejects `customVocabulary` and `mode: "SMART"` alongside diarization or word
+ * timestamps.
+ */
+export type GoogleTranscriptionOptions = {
+  readonly audioTranscriptionConfig?: {
+    readonly mode?: "VERBATIM" | "SMART" | (string & {})
+    readonly customVocabulary?: ReadonlyArray<string>
+    readonly languageCodes?: ReadonlyArray<string>
+  }
+} & Record<string, unknown>
+
+export type Request = TranscriptionRequestFor<GoogleTranscriptionOptions>
+
+// ---------------------------------------------------------------------------
+// 3. Streaming event schema
+// ---------------------------------------------------------------------------
+
+const Seconds = Schema.String.check(Schema.isPattern(/^\d+(\.\d+)?s$/)).pipe(
+  Schema.decodeTo(Schema.Number, {
+    decode: SchemaGetter.transform((value) => Number.parseFloat(value)),
+    encode: SchemaGetter.transform((value) => `${value}s`),
+  }),
+)
+
+const AudioTranscription = Schema.Struct({
+  text: Schema.String,
+  speakerLabel: Schema.optional(Schema.String),
+  words: Schema.optional(
+    Schema.Array(
+      Schema.Struct({
+        word: Schema.String,
+        startOffset: Seconds,
+        endOffset: Seconds,
+        speakerLabel: Schema.optional(Schema.String),
+      }),
+    ),
+  ),
+})
+
+const decodeChunk = MediaProtocol.decodeFrame(
+  ADAPTER,
+  NAME,
+  GeminiGenerateContent.chunk(Schema.Struct({ audioTranscription: Schema.optional(AudioTranscription) })),
+)
+
+// ---------------------------------------------------------------------------
+// 4. Parser state
+// ---------------------------------------------------------------------------
+
+interface State extends GeminiGenerateContent.Metadata {
+  readonly text: string
+  readonly segments: Array<TranscriptionSegment>
+  readonly words: Array<TranscriptionWord>
+}
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+const fromRequest = Effect.fn("GoogleTranscription.fromRequest")(function* (request: MediaProtocol.Addressed<Request>) {
+  // General Gemini models ignore `audioTranscriptionConfig` and answer the audio conversationally.
+  if (!request.model.id.includes("transcribe"))
+    return yield* ProviderShared.unsupportedOperation({
+      operation: "transcription.model",
+      provider: PROVIDER,
+      route: ADAPTER,
+      message: `${request.model.id} is not a transcription model; use a transcribe model such as gemini-3.5-transcribe`,
+    })
+  return MediaProtocol.json(
+    mergeJsonRecords(
+      {
+        contents: [{ role: "user", parts: [yield* GeminiGenerateContent.mediaPart(ADAPTER, request.audio)] }],
+        generationConfig: mergeJsonRecords(
+          {
+            audioTranscriptionConfig: {
+              languageCodes: request.language === undefined ? undefined : [request.language],
+              // Parts carry no offsets of their own, so segment times come from word offsets.
+              wordTimestamp:
+                request.diarize === true || request.timestamps === "word" || request.timestamps === "segment"
+                  ? true
+                  : undefined,
+              diarization: request.diarize === true ? true : undefined,
+            },
+          },
+          request.providerOptions,
+        ),
+      },
+      request.http?.body,
+    ) ?? {},
+  )
+})
+
+// ---------------------------------------------------------------------------
+// 6. Stream parsing
+// ---------------------------------------------------------------------------
+
+const turn = (part: Schema.Schema.Type<typeof AudioTranscription>) => {
+  const words = (part.words ?? []).map((word) => ({
+    text: word.word,
+    startSeconds: word.startOffset,
+    endSeconds: word.endOffset,
+    speaker: word.speakerLabel ?? part.speakerLabel,
+  }))
+  const first = words[0]
+  const last = words.at(-1)
+  return {
+    text: part.text,
+    words,
+    segment:
+      first === undefined || last === undefined
+        ? undefined
+        : {
+            text: part.text,
+            startSeconds: first.startSeconds,
+            endSeconds: last.endSeconds,
+            speaker: part.speakerLabel,
+          },
+  }
+}
+
+const step = Effect.fn("GoogleTranscription.step")(function* (state: State, frame: string) {
+  const chunk = yield* decodeChunk(frame)
+  const blocked = GeminiGenerateContent.blocked(NAME, chunk, frame)
+  if (blocked !== undefined) return yield* blocked
+  const turns = (chunk.candidates?.[0]?.content?.parts ?? []).flatMap((part) =>
+    part.audioTranscription === undefined ? [] : [turn(part.audioTranscription)],
+  )
+  const segments = turns.flatMap((item) => (item.segment === undefined ? [] : [item.segment]))
+  state.words.push(...turns.flatMap((item) => item.words))
+  state.segments.push(...segments)
+  // Each part is one whole speaker turn without surrounding whitespace, so turns join with a space.
+  const text = turns
+    .map((item) => item.text)
+    .filter((item) => item.length > 0)
+    .join(" ")
+  const delta = text.length === 0 || state.text.length === 0 ? text : ` ${text}`
+  const events: ReadonlyArray<TranscriptionEvent> = [
+    ...(delta.length === 0 ? [] : [TranscriptionTextDeltaEvent.make({ delta })]),
+    ...segments.map((segment) => TranscriptionSegmentEvent.make({ segment })),
+  ]
+  return [{ ...GeminiGenerateContent.track(state, chunk), text: state.text + delta }, events] as const
+})
+
+const finish = (state: State) => {
+  if (state.finishReason === undefined) return Effect.fail(MediaProtocol.incomplete(ADAPTER))
+  return Effect.succeed([
+    TranscriptionFinishEvent.make({
+      text: state.text,
+      segments: state.segments.length === 0 ? undefined : state.segments,
+      words: state.words.length === 0 ? undefined : state.words,
+      usage: GeminiGenerateContent.usage(state.usage),
+      providerMetadata: GeminiGenerateContent.providerMetadata(state),
+    }),
+  ])
+}
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+export const protocol = MediaProtocol.stream<Request, TranscriptionEvent, string, State>({
+  id: ADAPTER,
+  name: NAME,
+  unsupported: ["prompt", "speakers"],
+  body: { from: fromRequest },
+  frames: (bytes, context) => GeminiGenerateContent.frames(bytes, context.request.mode),
+  initial: () => ({ text: "", segments: [], words: [] }),
+  step,
+  finish,
+})
+
+export const model = (input: MediaRoute.ModelInput) =>
+  TranscriptionModel.fromRoute<GoogleTranscriptionOptions, string, State>(
+    {
+      id: ADAPTER,
+      provider: PROVIDER,
+      protocol,
+      baseURL: DEFAULT_BASE_URL,
+      path: ({ request }) => GeminiGenerateContent.path(request.model.id, request.mode),
+    },
+    input,
+  )
+
+export const GoogleTranscription = {
+  protocol,
+  model,
+} as const
@@ -0,0 +1,233 @@
+import { Duration, Effect, Schema } from "effect"
+import type { HttpClientResponse } from "effect/unstable/http"
+import type { Status } from "../generation.js"
+import { Media } from "../media.js"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
+import { VideoModel, VideoResponse, type VideoRequestFor } from "../video.js"
+import { ProviderShared, optionalArray } from "./shared.js"
+
+const ADAPTER = "google-video"
+const NAME = "Google Veo"
+const PROVIDER = ProviderID.make("google")
+export const DEFAULT_BASE_URL = "https://generativelanguage.googleapis.com/v1beta"
+/** Veo keeps generated files for two days; the asset carries that deadline so callers materialize in time. */
+const FILE_RETENTION = Duration.days(2)
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+export type GoogleVideoString<Known extends string> = Known | (string & {})
+
+/** Provider-native `parameters`. Common fields (`aspectRatio`, `resolution`, `durationSeconds`, `seed`) live on the request. */
+export type GoogleVideoOptions = {
+  readonly personGeneration?: GoogleVideoString<"allow_all" | "allow_adult" | "dont_allow">
+} & Record<string, unknown>
+
+export type Request = VideoRequestFor<GoogleVideoOptions>
+
+// ---------------------------------------------------------------------------
+// 2. Token and response schemas
+// ---------------------------------------------------------------------------
+
+/** The long-running operation name, e.g. `models/veo-3.1-generate-preview/operations/abc123`. */
+export const Token = Schema.Struct({ operation: Schema.String })
+export type Token = Schema.Schema.Type<typeof Token>
+
+const StartResponse = Schema.Struct({ name: Schema.String })
+
+const Operation = Schema.Struct({
+  done: Schema.optional(Schema.Boolean),
+  error: Schema.optional(Schema.Struct({ message: Schema.optional(Schema.String) })),
+  response: Schema.optional(
+    Schema.Struct({
+      generateVideoResponse: Schema.optional(
+        Schema.Struct({
+          generatedSamples: optionalArray(
+            Schema.Struct({
+              video: Schema.optional(
+                Schema.Struct({
+                  uri: Schema.optional(Schema.String),
+                  mimeType: Schema.optional(Schema.String),
+                }),
+              ),
+            }),
+          ),
+          raiMediaFilteredCount: Schema.optional(Schema.Number),
+          raiMediaFilteredReasons: optionalArray(Schema.String),
+        }),
+      ),
+    }),
+  ),
+  metadata: Schema.optional(Schema.Unknown),
+})
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+// Veo takes inline media only; a prior Veo output is `Media.url` with transient auth, so materialize it first.
+const inlineMedia = (asset: Media.Asset) =>
+  ProviderShared.requireInlineMedia(NAME, asset).pipe(
+    Effect.map((inline) => ({ inlineData: { mimeType: inline.mime, data: inline.base64 } })),
+  )
+
+const fromRequest = Effect.fn("GoogleVideo.fromRequest")(function* (request: Request) {
+  if (request.n !== undefined && request.n > 1)
+    return yield* ProviderShared.unsupportedOperation({
+      operation: "video.n",
+      provider: PROVIDER,
+      route: ADAPTER,
+      message: `${NAME} generates one video per request; call it once per video instead of n=${request.n}`,
+    })
+  if (request.audio === false)
+    return yield* ProviderShared.unsupportedOperation({
+      operation: "video.audio",
+      provider: PROVIDER,
+      route: ADAPTER,
+      message: `${NAME} always generates audio; audio: false cannot be honored`,
+    })
+  if (request.frames?.last !== undefined && request.frames.first === undefined)
+    return yield* ProviderShared.invalidRequest(`${NAME} requires frames.first when frames.last is set`)
+  const image = request.frames?.first === undefined ? undefined : yield* inlineMedia(request.frames.first)
+  const lastFrame = request.frames?.last === undefined ? undefined : yield* inlineMedia(request.frames.last)
+  const video = request.video === undefined ? undefined : yield* inlineMedia(request.video)
+  const referenceImages = yield* Effect.forEach(request.references ?? [], (asset) =>
+    inlineMedia(asset).pipe(Effect.map((image) => ({ image, referenceType: "asset" }))),
+  )
+  return MediaProtocol.json(
+    mergeJsonRecords(
+      {
+        instances: [
+          {
+            prompt: request.prompt,
+            image,
+            lastFrame,
+            referenceImages: referenceImages.length === 0 ? undefined : referenceImages,
+            video,
+          },
+        ],
+        parameters: mergeJsonRecords(
+          {
+            aspectRatio: request.aspectRatio,
+            resolution: request.resolution,
+            durationSeconds: request.durationSeconds,
+            negativePrompt: request.negativePrompt,
+            seed: request.seed,
+          },
+          request.providerOptions,
+        ),
+      },
+      request.http?.body,
+    ) ?? {},
+  )
+})
+
+// ---------------------------------------------------------------------------
+// 6. Response decoding
+// ---------------------------------------------------------------------------
+
+const decodeStart = MediaProtocol.decodeStarted(ADAPTER, NAME, StartResponse, (value) => ({
+  token: { operation: value.name },
+  snapshot: { id: value.name, status: "running" },
+}))
+
+// Operations carry no status string: not done is running, done with `error` failed, otherwise completed.
+const statusOf = (operation: typeof Operation.Type): Status => {
+  if (operation.done !== true) return "running"
+  return operation.error === undefined ? "completed" : "failed"
+}
+
+const decodeOperation = MediaProtocol.decodeJson(ADAPTER, NAME, Operation)
+
+const decodeStatus = Effect.fn("GoogleVideo.decodeStatus")(function* (
+  response: HttpClientResponse.HttpClientResponse,
+  context: MediaProtocol.PollContext<Token>,
+) {
+  const output = yield* decodeOperation(response)
+  return { id: context.token.operation, status: statusOf(output.value) }
+})
+
+const decodeResult = Effect.fn("GoogleVideo.decodeResult")(function* (
+  response: HttpClientResponse.HttpClientResponse,
+  context: MediaProtocol.PollContext<Token>,
+) {
+  const output = yield* decodeOperation(response)
+  const operation = output.value
+  const status = statusOf(operation)
+  if (status === "running")
+    return yield* output.invalid(`${NAME} operation ${context.token.operation} has not finished`)
+  if (status === "failed")
+    return yield* output.ended(
+      "failed",
+      `${NAME} operation failed${operation.error?.message === undefined ? "" : `: ${operation.error.message}`}`,
+    )
+  const generated = operation.response?.generateVideoResponse
+  // Downloads require the same API key as the poll; the asset carries it transiently and follows the redirect.
+  const videos = yield* Effect.forEach(
+    (generated?.generatedSamples ?? []).flatMap((sample) =>
+      sample.video?.uri === undefined ? [] : [{ uri: sample.video.uri, mimeType: sample.video.mimeType }],
+    ),
+    (video) =>
+      MediaProtocol.expiringUrl(video.uri, FILE_RETENTION, {
+        mediaType: video.mimeType ?? "video/mp4",
+        headers: context.auth,
+      }),
+  )
+  const reasons = generated?.raiMediaFilteredReasons ?? []
+  const notices = reasons.map((reason) => ({
+    type: "filtered" as const,
+    message: `${NAME} filtered media: ${reason}`,
+    providerMetadata: { google: { raiMediaFilteredReason: reason } },
+  }))
+  if (videos.length === 0 && (reasons.length > 0 || (generated?.raiMediaFilteredCount ?? 0) > 0))
+    return yield* output.contentPolicy(
+      `${NAME} filtered every video${reasons.length === 0 ? "" : `: ${reasons.join("; ")}`}`,
+    )
+  if (videos.length === 0) return yield* output.invalid(`${NAME} operation completed without any video`)
+  return new VideoResponse({
+    videos,
+    notices: notices.length === 0 ? undefined : notices,
+    providerMetadata: {
+      google: {
+        operation: context.token.operation,
+        raiMediaFilteredCount: generated?.raiMediaFilteredCount,
+        metadata: operation.metadata,
+      },
+    },
+  })
+})
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+const operationPath = (token: Token) => `/${token.operation}`
+
+export const protocol = MediaProtocol.queued<Request, VideoResponse, Token>({
+  id: ADAPTER,
+  name: NAME,
+  token: Token,
+  start: { body: { from: fromRequest }, decode: decodeStart },
+  status: { path: operationPath, decode: decodeStatus },
+  result: { path: operationPath, decode: decodeResult },
+})
+
+export const model = (input: MediaRoute.ModelInput) =>
+  VideoModel.fromRoute<GoogleVideoOptions, Token>(
+    {
+      id: ADAPTER,
+      provider: PROVIDER,
+      protocol,
+      baseURL: DEFAULT_BASE_URL,
+      path: ({ request }) => `/models/${request.model.id}:predictLongRunning`,
+    },
+    input,
+  )
+
+export const GoogleVideo = {
+  protocol,
+  model,
+} as const
@@ -4,7 +4,7 @@ import { ImageModel, ImageResponse, type ImageRequestFor } from "../image.js"
 import { Media } from "../media.js"
 import { MediaProtocol } from "../route/media-protocol.js"
 import { MediaRoute } from "../route/media.js"
-import { ProviderID, mergeJsonRecords, type AIError } from "../schema/index.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
 import { JsonObject, ProviderShared, optionalNull } from "./shared.js"
 import { MediaInput } from "./utils/media-input.js"

@@ -70,13 +70,9 @@ const Response = Schema.Struct({

 const isEdit = (request: Request) => (request.images?.length ?? 0) > 0

-const reference = (asset: Media.Asset): Effect.Effect<Record<string, unknown>, AIError> => {
-  const inline = asset.inline()
-  if (inline) return Effect.succeed({ image_url: inline.dataUrl })
-  const url = ProviderShared.mediaUrl(asset)
-  if (url) return Effect.succeed({ image_url: url })
-  return Effect.fail(ProviderShared.invalidRequest(`${NAME} accepts image bytes and URLs`))
-}
+// Meta has no file handles: refs are rejected even when they name this provider.
+const reference = (asset: Media.Asset) =>
+  ProviderShared.mediaReference(asset, undefined, NAME).pipe(Effect.map((item) => ({ image_url: item.value })))

 const fromRequest = Effect.fn("MetaImages.fromRequest")(function* (request: Request) {
  const images = yield* Effect.forEach(request.images ?? [], reference)
@@ -130,6 +130,10 @@ const OpenAIChatUserContent = Schema.Union([
    type: Schema.Literal("image_url"),
    image_url: Schema.Struct({ url: Schema.String }),
  }),
+  Schema.Struct({
+    type: Schema.Literal("file"),
+    file: Schema.Struct({ filename: Schema.String, file_data: Schema.String }),
+  }),
 ])

 const OpenAIChatMessage = Schema.Union([
@@ -360,6 +364,15 @@ const lowerToolCall = (part: ToolCallPart, options: LoweringOptions): OpenAIChat
 })

 const lowerMedia = Effect.fn("OpenAIChat.lowerMedia")(function* (part: MediaPart) {
+  // Chat Completions accepts PDFs, and no other documents, as inline `file` parts; file URLs are not supported.
+  if (part.media.mediaType.toLowerCase() === "application/pdf")
+    return {
+      type: "file" as const,
+      file: {
+        filename: part.filename ?? "document.pdf",
+        file_data: (yield* ProviderShared.requireInlineMedia("OpenAI Chat", part.media)).dataUrl,
+      },
+    }
  if (part.media.kind !== "image")
    return yield* ProviderShared.invalidRequest(`OpenAI Chat does not support media type ${part.media.mediaType}`)
  const url =
@@ -489,7 +502,7 @@ const lowerToolMessages = Effect.fn("OpenAIChat.lowerToolMessages")(function* (
  options: LoweringOptions,
 ) {
  const messages: OpenAIChatMessage[] = []
-  const images: Array<Schema.Schema.Type<typeof OpenAIChatUserContent>> = []
+  const attachments: Array<Schema.Schema.Type<typeof OpenAIChatUserContent>> = []
  for (const part of message.content) {
    if (!ProviderShared.supportsContent(part, ["tool-result"]))
      return yield* ProviderShared.unsupportedContent("OpenAI Chat", "tool", ["tool-result"])
@@ -511,9 +524,9 @@ const lowerToolMessages = Effect.fn("OpenAIChat.lowerToolMessages")(function* (
      cache_control: options.cacheControl?.(part.cache),
    })
    const files = content.filter((item) => item.type === "file")
-    images.push(...(yield* Effect.forEach(files, (item) => lowerMedia(ProviderShared.toolFileMedia(item)))))
+    attachments.push(...(yield* Effect.forEach(files, (item) => lowerMedia(ProviderShared.toolFileMedia(item)))))
  }
-  return { messages, images }
+  return { messages, attachments }
 })

 const lowerMessage = Effect.fn("OpenAIChat.lowerMessage")(function* (
@@ -574,21 +587,21 @@ const lowerMessages = Effect.fn("OpenAIChat.lowerMessages")(function* (request:
    if (requireAssistantAfterTool && messages.at(-1)?.role === "tool")
      messages.push({ role: "assistant", content: "Done." })
  }
-  const pendingImages: Array<Schema.Schema.Type<typeof OpenAIChatUserContent>> = []
-  const flushImages = () => {
-    if (pendingImages.length === 0) return
+  const pendingAttachments: Array<Schema.Schema.Type<typeof OpenAIChatUserContent>> = []
+  const flushAttachments = () => {
+    if (pendingAttachments.length === 0) return
    bridgeTools()
-    messages.push({ role: "user", content: pendingImages.splice(0) })
+    messages.push({ role: "user", content: pendingAttachments.splice(0) })
  }
  for (const message of request.messages) {
    if (message.role === "user") bridgeTools()
    if (message.role === "system") {
      const part = yield* ProviderShared.wrappedSystemUpdate("OpenAI Chat", message)
-      if (pendingImages.length > 0) {
+      if (pendingAttachments.length > 0) {
        messages.push({
          role: "user",
          content: [
-            ...pendingImages.splice(0),
+            ...pendingAttachments.splice(0),
            { type: "text", text: part.text, cache_control: options.cacheControl?.(part.cache) },
          ],
        })
@@ -632,13 +645,13 @@ const lowerMessages = Effect.fn("OpenAIChat.lowerMessages")(function* (request:
    if (message.role === "tool") {
      const lowered = yield* lowerToolMessages(message, lowering)
      messages.push(...lowered.messages)
-      pendingImages.push(...lowered.images)
+      pendingAttachments.push(...lowered.attachments)
      continue
    }
-    flushImages()
+    flushAttachments()
    messages.push(...(yield* lowerMessage(message, reasoningField, requireReasoning, lowering)))
  }
-  flushImages()
+  flushAttachments()
  return messages
 })

@@ -4,7 +4,7 @@ import { ImageModel, ImageResponse, type ImageRequestFor } from "../image.js"
 import { Media } from "../media.js"
 import { MediaProtocol } from "../route/media-protocol.js"
 import { MediaRoute } from "../route/media.js"
-import { ProviderID, mergeJsonRecords, type AIError } from "../schema/index.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
 import { ProviderShared } from "./shared.js"
 import { MediaInput } from "./utils/media-input.js"

@@ -72,23 +72,10 @@ const isEdit = (request: Request) => (request.images?.length ?? 0) > 0

 const isInline = (asset: Media.Asset) => asset.inline() !== undefined

-const blob = (data: Uint8Array, mediaType: string) => {
-  const buffer = new ArrayBuffer(data.byteLength)
-  new Uint8Array(buffer).set(data)
-  return new Blob([buffer], { type: mediaType })
-}
-
-const reference = (asset: Media.Asset): Effect.Effect<Record<string, unknown>, AIError> => {
-  const inline = asset.inline()
-  if (inline) return Effect.succeed({ image_url: inline.dataUrl })
-  const url = ProviderShared.mediaUrl(asset)
-  if (url) return Effect.succeed({ image_url: url })
-  const id = MediaInput.refID(asset, PROVIDER)
-  if (id) return Effect.succeed({ file_id: id })
-  return Effect.fail(
-    ProviderShared.invalidRequest("OpenAI Images accepts image URLs, data URLs, bytes, and OpenAI file IDs"),
+const reference = (asset: Media.Asset) =>
+  ProviderShared.mediaReference(asset, PROVIDER, NAME).pipe(
+    Effect.map((item) => (item.type === "ref" ? { file_id: item.value } : { image_url: item.value })),
  )
-}

 const fromRequest = Effect.fn("OpenAIImages.fromRequest")(function* (request: Request) {
  const images = request.images ?? []
@@ -111,9 +98,11 @@ const fromRequest = Effect.fn("OpenAIImages.fromRequest")(function* (request: Re
      form.append(key, typeof value === "string" ? value : ProviderShared.encodeJson(value))
    })
    const uploads = yield* Effect.forEach(images, (image) => MediaInput.inlineBytes(ADAPTER, image))
-    uploads.forEach((data, index) => form.append("image[]", blob(data, images[index].mediaType), `image-${index}`))
+    uploads.forEach((data, index) =>
+      form.append("image[]", MediaInput.blob(data, images[index].mediaType), `image-${index}`),
+    )
    if (mask !== undefined)
-      form.append("mask", blob(yield* MediaInput.inlineBytes(ADAPTER, mask), mask.mediaType), "mask")
+      form.append("mask", MediaInput.blob(yield* MediaInput.inlineBytes(ADAPTER, mask), mask.mediaType), "mask")
    return MediaProtocol.multipart(form)
  }

@@ -137,6 +126,7 @@ const fromRequest = Effect.fn("OpenAIImages.fromRequest")(function* (request: Re
 // ---------------------------------------------------------------------------

 const requestedFormat = (body: MediaProtocol.Body) => {
+  if (body.type === "binary") return undefined
  const value = body.type === "json" ? body.value.output_format : body.value.get("output_format")
  return typeof value === "string" ? value : undefined
 }
@@ -0,0 +1,141 @@
+import { Effect, Schema } from "effect"
+import { Framing } from "../route/framing.js"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { ProviderID, mergeJsonRecords, type MediaUsage } from "../schema/index.js"
+import { SpeechModel, type SpeechEvent, type SpeechRequestFor } from "../speech.js"
+import { SpeechStream } from "./utils/speech-stream.js"
+
+const ADAPTER = "openai-speech"
+const NAME = "OpenAI Speech"
+const PROVIDER = ProviderID.make("openai")
+export const DEFAULT_BASE_URL = "https://api.openai.com/v1"
+export const PATH = "/audio/speech"
+/** `pcm` is raw 24 kHz, 16-bit signed little-endian mono samples without a header. */
+const PCM_SAMPLE_RATE = 24000
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+export type OpenAISpeechOptions = Record<string, unknown>
+
+export type Request = SpeechRequestFor<OpenAISpeechOptions>
+
+// ---------------------------------------------------------------------------
+// 3. Streaming event schema
+// ---------------------------------------------------------------------------
+
+const SpeechStreamEvent = Schema.Union([
+  Schema.Struct({ type: Schema.Literal("speech.audio.delta"), audio: Schema.Uint8ArrayFromBase64 }),
+  Schema.Struct({
+    type: Schema.Literal("speech.audio.done"),
+    usage: Schema.optional(
+      Schema.Struct({
+        input_tokens: Schema.optional(Schema.Number),
+        output_tokens: Schema.optional(Schema.Number),
+        total_tokens: Schema.optional(Schema.Number),
+      }),
+    ),
+  }),
+])
+
+const decodeEvent = MediaProtocol.decodeFrame(ADAPTER, NAME, SpeechStreamEvent)
+
+// ---------------------------------------------------------------------------
+// 4. Parser state
+// ---------------------------------------------------------------------------
+
+interface State extends SpeechStream.Audio {
+  readonly done: boolean
+  readonly usage?: MediaUsage
+}
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+// `sse` is not supported for `tts-1` or `tts-1-hd`; those models stream the raw audio body instead.
+const supportsSse = (model: string) => !/^tts-1(-hd)?(-|$)/.test(model)
+
+const fromRequest = Effect.fn("OpenAISpeech.fromRequest")(function* (request: MediaProtocol.Addressed<Request>) {
+  return MediaProtocol.json(
+    mergeJsonRecords(
+      {
+        model: request.model.id,
+        input: request.text,
+        voice: request.voice,
+        instructions: request.instructions,
+        response_format: request.format,
+        speed: request.speed,
+        stream_format: request.mode === "stream" && supportsSse(request.model.id) ? "sse" : undefined,
+      },
+      request.providerOptions,
+      request.http?.body,
+    ) ?? {},
+  )
+})
+
+// ---------------------------------------------------------------------------
+// 6. Stream parsing
+// ---------------------------------------------------------------------------
+
+const isSse = (body: MediaProtocol.Body) => body.type === "json" && body.value.stream_format === "sse"
+
+const onEvent = Effect.fn("OpenAISpeech.onEvent")(function* (state: State, frame: string) {
+  const event = yield* decodeEvent(frame)
+  if (event.type === "speech.audio.delta") return SpeechStream.delta(state, event.audio)
+  const usage = event.usage
+  return [
+    {
+      ...state,
+      done: true,
+      usage:
+        usage === undefined
+          ? undefined
+          : {
+              type: "tokens" as const,
+              input: usage.input_tokens,
+              output: usage.output_tokens,
+              total: usage.total_tokens,
+              details: { openai: usage },
+            },
+    },
+    [],
+  ] as const
+})
+
+const finish = (state: State, context: MediaProtocol.ResponseContext<Request>) => {
+  if (isSse(context.body) && !state.done) return Effect.fail(MediaProtocol.incomplete(ADAPTER))
+  const format = context.request.format ?? "mp3"
+  return SpeechStream.finish(ADAPTER, state, {
+    ...(format === "pcm" ? SpeechStream.pcm("pcm_s16le", PCM_SAMPLE_RATE) : SpeechStream.container(format)),
+    usage: state.usage,
+  })
+}
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+export const protocol = MediaProtocol.stream<Request, SpeechEvent, string | Uint8Array, State>({
+  id: ADAPTER,
+  name: NAME,
+  unsupported: ["language", "timestamps"],
+  body: { from: fromRequest },
+  frames: (bytes, context) => (isSse(context.body) ? Framing.sse.frame(bytes) : bytes),
+  initial: () => ({ chunks: [], done: false }),
+  step: SpeechStream.step(onEvent),
+  finish,
+})
+
+export const model = (input: MediaRoute.ModelInput) =>
+  SpeechModel.fromRoute<OpenAISpeechOptions, string | Uint8Array, State>(
+    { id: ADAPTER, provider: PROVIDER, protocol, baseURL: DEFAULT_BASE_URL, path: PATH },
+    input,
+  )
+
+export const OpenAISpeech = {
+  protocol,
+  model,
+} as const
@@ -0,0 +1,277 @@
+import { Effect, Schema, Stream } from "effect"
+import { Framing } from "../route/framing.js"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { ProviderID, mergeJsonRecords, type MediaUsage } from "../schema/index.js"
+import {
+  TranscriptionFinishEvent,
+  TranscriptionModel,
+  TranscriptionSegmentEvent,
+  TranscriptionTextDeltaEvent,
+  type TranscriptionEvent,
+  type TranscriptionRequestFor,
+  type TranscriptionSegment,
+} from "../transcription.js"
+import { mediaTypeExtension } from "../utils/media-type.js"
+import { ProviderShared } from "./shared.js"
+import { MediaInput } from "./utils/media-input.js"
+
+const ADAPTER = "openai-transcription"
+const NAME = "OpenAI Transcription"
+const PROVIDER = ProviderID.make("openai")
+export const DEFAULT_BASE_URL = "https://api.openai.com/v1"
+export const PATH = "/audio/transcriptions"
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+export type OpenAITranscriptionOptions = {
+  readonly chunking_strategy?:
+    | "auto"
+    | {
+        readonly type: "server_vad"
+        readonly prefix_padding_ms?: number
+        readonly silence_duration_ms?: number
+        readonly threshold?: number
+      }
+  readonly include?: ReadonlyArray<"logprobs">
+  readonly keywords?: ReadonlyArray<string>
+  readonly known_speaker_names?: ReadonlyArray<string>
+  readonly known_speaker_references?: ReadonlyArray<string>
+  readonly temperature?: number
+} & Record<string, unknown>
+
+export type Request = TranscriptionRequestFor<OpenAITranscriptionOptions>
+
+// ---------------------------------------------------------------------------
+// 3. Streaming event schema
+// ---------------------------------------------------------------------------
+
+const Segment = Schema.Struct({
+  text: Schema.String,
+  start: Schema.Number,
+  end: Schema.Number,
+  speaker: Schema.optional(Schema.String),
+})
+
+const Usage = Schema.Union([
+  Schema.Struct({
+    type: Schema.Literal("tokens"),
+    input_tokens: Schema.optional(Schema.Number),
+    output_tokens: Schema.optional(Schema.Number),
+    total_tokens: Schema.optional(Schema.Number),
+  }),
+  Schema.Struct({ type: Schema.Literal("duration"), seconds: Schema.Number }),
+])
+
+const transcriptFields = {
+  text: Schema.String,
+  language: Schema.optional(Schema.String),
+  languages: Schema.optional(Schema.Array(Schema.Struct({ code: Schema.String }))),
+  duration: Schema.optional(Schema.Number),
+  segments: Schema.optional(Schema.Array(Segment)),
+  words: Schema.optional(
+    Schema.Array(Schema.Struct({ word: Schema.String, start: Schema.Number, end: Schema.Number })),
+  ),
+  usage: Schema.optional(Usage),
+}
+
+const Event = Schema.Union([
+  Schema.Struct({ type: Schema.Literal("transcript.text.delta"), delta: Schema.String }),
+  Schema.Struct({ type: Schema.Literal("transcript.text.segment"), ...Segment.fields }),
+  Schema.Struct({ type: Schema.Literal("transcript.text.done"), ...transcriptFields }),
+])
+const Transcript = Schema.Struct(transcriptFields)
+type Transcript = Schema.Schema.Type<typeof Transcript>
+
+const decodeEvent = MediaProtocol.decodeFrame(ADAPTER, NAME, Event)
+const decodeTranscript = MediaProtocol.decodeFrame(ADAPTER, NAME, Transcript)
+
+type Frame = string | { readonly document: string }
+
+// ---------------------------------------------------------------------------
+// 4. Parser state
+// ---------------------------------------------------------------------------
+
+interface State {
+  readonly segments: Array<TranscriptionSegment>
+  readonly transcript?: Transcript
+}
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+interface Capabilities {
+  readonly stream: boolean
+  readonly timestamps: ReadonlyArray<"segment" | "word">
+  readonly diarize: boolean
+  readonly languageField: "language" | "languages"
+}
+
+const TRANSCRIBE: Capabilities = { stream: true, timestamps: [], diarize: false, languageField: "language" }
+
+const capabilities = (model: string): Capabilities => {
+  if (model.startsWith("whisper")) return { ...TRANSCRIBE, stream: false, timestamps: ["segment", "word"] }
+  if (model.includes("diarize")) return { ...TRANSCRIBE, timestamps: ["segment"], diarize: true }
+  // `gpt-transcribe` replaces `language` with `languages[]` and rejects both together.
+  if (model.startsWith("gpt-transcribe")) return { ...TRANSCRIBE, languageField: "languages" }
+  return TRANSCRIBE
+}
+
+const unsupported = (operation: string, message: string) =>
+  Effect.fail(ProviderShared.unsupportedOperation({ operation, provider: PROVIDER, route: ADAPTER, message }))
+
+const validate = (request: MediaProtocol.Addressed<Request>, model: Capabilities) => {
+  const id = request.model.id
+  if (request.mode === "stream" && !model.stream)
+    return unsupported("media.stream", `${id} does not stream; use Transcription.generate`)
+  if (request.diarize === true && !model.diarize)
+    return unsupported("media.diarize", `${id} does not diarize; use gpt-4o-transcribe-diarize`)
+  if (request.prompt !== undefined && model.diarize)
+    return unsupported("media.prompt", `${id} does not accept a prompt`)
+  if (
+    request.timestamps === undefined ||
+    request.timestamps === "none" ||
+    model.timestamps.includes(request.timestamps)
+  )
+    return Effect.void
+  return unsupported("media.timestamps", `${id} does not return ${request.timestamps} timestamps`)
+}
+
+const RESERVED_FORM_FIELDS = new Set([
+  "file",
+  "model",
+  "prompt",
+  "language",
+  "response_format",
+  "timestamp_granularities",
+  "stream",
+])
+
+const appendField = (form: FormData, key: string, value: unknown) => {
+  if (Array.isArray(value)) return value.forEach((item) => form.append(`${key}[]`, String(item)))
+  form.append(key, typeof value === "object" && value !== null ? ProviderShared.encodeJson(value) : String(value))
+}
+
+const fromRequest = Effect.fn("OpenAITranscription.fromRequest")(function* (request: MediaProtocol.Addressed<Request>) {
+  const model = capabilities(request.model.id)
+  yield* validate(request, model)
+  // The API detects the audio format from the upload's filename extension.
+  const extension = mediaTypeExtension(request.audio.mediaType)
+  if (extension === undefined)
+    return yield* ProviderShared.invalidRequest(
+      `${NAME} cannot name a ${request.audio.mediaType} upload; send mp3, mp4, m4a, wav, webm, ogg, or flac audio`,
+    )
+  const audio = yield* MediaInput.inlineBytes(ADAPTER, request.audio)
+  const responseFormat = model.diarize
+    ? "diarized_json"
+    : request.timestamps === undefined || request.timestamps === "none"
+      ? undefined
+      : "verbose_json"
+  const native = Object.entries(mergeJsonRecords(request.providerOptions, request.http?.body) ?? {}).filter(
+    ([key]) => !RESERVED_FORM_FIELDS.has(key),
+  )
+  const fields = mergeJsonRecords(
+    {
+      model: request.model.id,
+      language: model.languageField === "language" ? request.language : undefined,
+      languages: model.languageField === "languages" && request.language !== undefined ? [request.language] : undefined,
+      prompt: request.prompt,
+      response_format: responseFormat,
+      timestamp_granularities: responseFormat === "verbose_json" ? [request.timestamps] : undefined,
+      // Diarizing audio longer than 30 seconds requires a chunking strategy.
+      chunking_strategy: model.diarize ? "auto" : undefined,
+      stream: request.mode === "stream" ? true : undefined,
+    },
+    Object.fromEntries(native),
+  )
+  const form = new FormData()
+  form.append("file", MediaInput.blob(audio, request.audio.mediaType), `audio.${extension}`)
+  Object.entries(fields ?? {}).forEach(([key, value]) => appendField(form, key, value))
+  return MediaProtocol.multipart(form)
+})
+
+// ---------------------------------------------------------------------------
+// 6. Stream parsing
+// ---------------------------------------------------------------------------
+
+const segment = (value: Schema.Schema.Type<typeof Segment>): TranscriptionSegment => ({
+  text: value.text.trim(),
+  startSeconds: value.start,
+  endSeconds: value.end,
+  speaker: value.speaker,
+})
+
+const onEvent = Effect.fn("OpenAITranscription.onEvent")(function* (state: State, frame: string) {
+  const event = yield* decodeEvent(frame)
+  if (event.type === "transcript.text.done") return [{ ...state, transcript: event }, []] as const
+  if (event.type === "transcript.text.delta")
+    return [state, event.delta.length === 0 ? [] : [TranscriptionTextDeltaEvent.make({ delta: event.delta })]] as const
+  const next = segment(event)
+  state.segments.push(next)
+  return [state, [TranscriptionSegmentEvent.make({ segment: next })]] as const
+})
+
+const step = (state: State, frame: Frame) =>
+  typeof frame === "string"
+    ? onEvent(state, frame)
+    : decodeTranscript(frame.document).pipe(Effect.map((transcript) => [{ ...state, transcript }, []] as const))
+
+const usage = (value: Transcript["usage"]): MediaUsage | undefined => {
+  if (value === undefined) return undefined
+  if (value.type === "duration") return { type: "seconds", seconds: value.seconds }
+  return {
+    type: "tokens",
+    input: value.input_tokens,
+    output: value.output_tokens,
+    total: value.total_tokens,
+    details: { openai: value },
+  }
+}
+
+const finish = (state: State) => {
+  const transcript = state.transcript
+  if (transcript === undefined) return Effect.fail(MediaProtocol.incomplete(ADAPTER))
+  const segments = transcript.segments?.map(segment) ?? state.segments
+  return Effect.succeed([
+    TranscriptionFinishEvent.make({
+      text: transcript.text,
+      segments: segments.length === 0 ? undefined : segments,
+      words: transcript.words?.map((word) => ({ text: word.word, startSeconds: word.start, endSeconds: word.end })),
+      language: (transcript.language ?? transcript.languages?.[0]?.code)?.toLowerCase(),
+      durationSeconds: transcript.duration,
+      usage: usage(transcript.usage),
+    }),
+  ])
+}
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+export const protocol = MediaProtocol.stream<Request, TranscriptionEvent, Frame, State>({
+  id: ADAPTER,
+  name: NAME,
+  unsupported: ["speakers"],
+  body: { from: fromRequest },
+  frames: (bytes, context) =>
+    context.request.mode === "stream"
+      ? Framing.sse.frame(bytes)
+      : Framing.document.frame(bytes).pipe(Stream.map((document) => ({ document }))),
+  initial: () => ({ segments: [] }),
+  step,
+  finish,
+})
+
+export const model = (input: MediaRoute.ModelInput) =>
+  TranscriptionModel.fromRoute<OpenAITranscriptionOptions, Frame, State>(
+    { id: ADAPTER, provider: PROVIDER, protocol, baseURL: DEFAULT_BASE_URL, path: PATH },
+    input,
+  )
+
+export const OpenAITranscription = {
+  protocol,
+  model,
+} as const
@@ -0,0 +1,204 @@
+import { Duration, Effect, Schema } from "effect"
+import type { HttpClientResponse } from "effect/unstable/http"
+import type { Status } from "../generation.js"
+import { Media } from "../media.js"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
+import { VideoModel, VideoResponse, type VideoRequestFor } from "../video.js"
+import { ProviderShared, optionalArray, optionalNull } from "./shared.js"
+
+const ADAPTER = "runway-video"
+const NAME = "Runway"
+const PROVIDER = ProviderID.make("runway")
+export const DEFAULT_BASE_URL = "https://api.dev.runwayml.com/v1"
+/** Every Runway request must pin the API version. */
+export const API_VERSION = "2024-11-06"
+export const TEXT_TO_VIDEO_PATH = "/text_to_video"
+export const IMAGE_TO_VIDEO_PATH = "/image_to_video"
+export const VIDEO_TO_VIDEO_PATH = "/video_to_video"
+export const TASKS_PATH = "/tasks"
+/** Output URLs are valid for 24–48 hours; the asset carries the conservative bound. */
+const OUTPUT_RETENTION = Duration.hours(24)
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+export type RunwayVideoString<Known extends string> = Known | (string & {})
+
+/**
+ * Provider-native options. Common fields lower to Runway's names: `aspectRatio` → `ratio` (Runway expects pixel
+ * ratios such as `1280:720` for most models), `durationSeconds` → `duration`, `audio`, `negativePrompt`,
+ * `resolution`, `references`, and `frames` → `promptImage`.
+ */
+export type RunwayVideoOptions = {
+  readonly contentModeration?: { readonly publicFigureThreshold?: RunwayVideoString<"auto" | "low"> }
+  readonly outputFormat?: RunwayVideoString<"mp4" | "prores" | "png_sequence">
+} & Record<string, unknown>
+
+export type Request = VideoRequestFor<RunwayVideoOptions>
+
+// ---------------------------------------------------------------------------
+// 2. Token and response schemas
+// ---------------------------------------------------------------------------
+
+export const Token = Schema.Struct({ taskID: Schema.String })
+export type Token = Schema.Schema.Type<typeof Token>
+
+const Cost = Schema.Struct({ credits: Schema.Number })
+
+const StartResponse = Schema.Struct({ id: Schema.String })
+
+const Task = Schema.Struct({
+  status: Schema.String,
+  progress: optionalNull(Schema.Number),
+  output: optionalArray(Schema.String),
+  failure: optionalNull(Schema.String),
+  failureCode: optionalNull(Schema.String),
+  cost: Schema.optional(Cost),
+  estimatedCost: Schema.optional(Cost),
+})
+
+const STATUS = {
+  PENDING: "queued",
+  THROTTLED: "queued",
+  RUNNING: "running",
+  SUCCEEDED: "completed",
+  FAILED: "failed",
+  CANCELLED: "cancelled",
+} as const satisfies Record<string, Status>
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+// Runway accepts HTTPS URLs, `runway://` upload URIs, and data URIs, all as one string.
+const mediaUri = (asset: Media.Asset) =>
+  ProviderShared.mediaReference(asset, PROVIDER, NAME).pipe(Effect.map((reference) => reference.value))
+
+const fromRequest = Effect.fn("RunwayVideo.fromRequest")(function* (request: Request) {
+  const first = request.frames?.first === undefined ? undefined : yield* mediaUri(request.frames.first)
+  const last = request.frames?.last === undefined ? undefined : yield* mediaUri(request.frames.last)
+  const promptImage = [
+    ...(first === undefined ? [] : [{ uri: first, position: "first" }]),
+    ...(last === undefined ? [] : [{ uri: last, position: "last" }]),
+  ]
+  const videoUri = request.video === undefined ? undefined : yield* mediaUri(request.video)
+  const references = yield* Effect.forEach(request.references ?? [], (asset) =>
+    mediaUri(asset).pipe(Effect.map((uri) => ({ uri }))),
+  )
+  return MediaProtocol.json(
+    mergeJsonRecords(
+      {
+        model: request.model.id,
+        promptText: request.prompt,
+        promptImage: promptImage.length === 0 ? undefined : promptImage,
+        videoUri,
+        references: references.length === 0 ? undefined : references,
+        ratio: request.aspectRatio,
+        duration: request.durationSeconds,
+        resolution: request.resolution,
+        audio: request.audio,
+        negativePrompt: request.negativePrompt,
+        seed: request.seed,
+      },
+      request.providerOptions,
+      request.http?.body,
+    ) ?? {},
+  )
+})
+
+// ---------------------------------------------------------------------------
+// 6. Response decoding
+// ---------------------------------------------------------------------------
+
+const decodeStart = MediaProtocol.decodeStarted(ADAPTER, NAME, StartResponse, (value) => ({
+  token: { taskID: value.id },
+  snapshot: { id: value.id, status: "queued" },
+}))
+
+const decodeTask = MediaProtocol.decodeJson(ADAPTER, NAME, Task)
+
+const decodeStatus = Effect.fn("RunwayVideo.decodeStatus")(function* (
+  response: HttpClientResponse.HttpClientResponse,
+  context: MediaProtocol.PollContext<Token>,
+) {
+  const output = yield* decodeTask(response)
+  const status = yield* MediaProtocol.status(STATUS, output.value.status, output)
+  return { id: context.token.taskID, status, progress: output.value.progress ?? undefined }
+})
+
+const decodeResult = Effect.fn("RunwayVideo.decodeResult")(function* (
+  response: HttpClientResponse.HttpClientResponse,
+  context: MediaProtocol.PollContext<Token>,
+) {
+  const output = yield* decodeTask(response)
+  const task = output.value
+  const status = yield* MediaProtocol.status(STATUS, task.status, output)
+  if (status === "failed") {
+    const code = task.failureCode ?? undefined
+    const message = `${NAME} task failed${code === undefined ? "" : ` (${code})`}${task.failure ? `: ${task.failure}` : ""}`
+    // Runway failure codes are dotted paths; every moderation outcome carries a SAFETY segment.
+    if (code !== undefined && /(^|\.)SAFETY(\.|$)/.test(code)) return yield* output.contentPolicy(message)
+    return yield* output.ended("failed", message)
+  }
+  if (status === "cancelled")
+    return yield* output.ended("cancelled", `${NAME} task ${context.token.taskID} was cancelled`)
+  if (status !== "completed") return yield* output.invalid(`${NAME} task ${context.token.taskID} has not finished`)
+  const urls = task.output ?? []
+  if (urls.length === 0) return yield* output.invalid(`${NAME} task succeeded without any output`)
+  return new VideoResponse({
+    videos: yield* Effect.forEach(urls, (url) =>
+      MediaProtocol.expiringUrl(url, OUTPUT_RETENTION, { mediaType: "video/mp4" }),
+    ),
+    usage: task.cost === undefined ? undefined : { type: "credits", credits: task.cost.credits },
+    providerMetadata: {
+      runway: {
+        taskId: context.token.taskID,
+        estimatedCredits: task.estimatedCost?.credits,
+      },
+    },
+  })
+})
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+const taskPath = (token: Token) => `${TASKS_PATH}/${token.taskID}`
+
+export const protocol = MediaProtocol.queued<Request, VideoResponse, Token>({
+  id: ADAPTER,
+  name: NAME,
+  token: Token,
+  unsupported: ["n"],
+  start: { body: { from: fromRequest }, decode: decodeStart },
+  status: { path: taskPath, decode: decodeStatus },
+  result: { path: taskPath, decode: decodeResult },
+  cancel: { method: "DELETE", path: taskPath },
+})
+
+const startPath = (request: Request) => {
+  if (request.video !== undefined) return VIDEO_TO_VIDEO_PATH
+  if (request.frames?.first !== undefined || request.frames?.last !== undefined) return IMAGE_TO_VIDEO_PATH
+  return TEXT_TO_VIDEO_PATH
+}
+
+export const model = (input: MediaRoute.ModelInput) =>
+  VideoModel.fromRoute<RunwayVideoOptions, Token>(
+    {
+      id: ADAPTER,
+      provider: PROVIDER,
+      protocol,
+      baseURL: DEFAULT_BASE_URL,
+      headers: { "X-Runway-Version": API_VERSION },
+      path: ({ request }) => startPath(request),
+    },
+    input,
+  )
+
+export const RunwayVideo = {
+  protocol,
+  model,
+} as const
@@ -192,6 +192,28 @@ export const inlineRequired = (route: string, asset: Media.Asset) =>
 /** The remote URL of a `url` asset, for protocols that accept `http(s)` references natively. */
 export const mediaUrl = (asset: Media.Asset) => (asset.source.type === "url" ? asset.source.url : undefined)

+export type MediaReference = { readonly type: "dataUrl" | "url" | "ref"; readonly value: string }
+
+/**
+ * The one string a provider can address an asset by: inline payloads as a data URL, `url` sources as their URL, and
+ * this provider's own `ref` as its id. Other providers' refs are never forwarded and fail typed; omit `provider` for
+ * APIs with no file handles at all.
+ */
+export const mediaReference = (
+  asset: Media.Asset,
+  provider: ProviderID | undefined,
+  label: string,
+): Effect.Effect<MediaReference, AIError> => {
+  const inline = asset.inline()
+  if (inline) return Effect.succeed({ type: "dataUrl", value: inline.dataUrl })
+  const url = mediaUrl(asset)
+  if (url) return Effect.succeed({ type: "url", value: url })
+  if (provider !== undefined && asset.source.type === "ref" && asset.source.provider === provider)
+    return Effect.succeed({ type: "ref", value: asset.source.id })
+  const accepted = provider === undefined ? "" : `, and ${provider} references`
+  return Effect.fail(invalidRequest(`${label} accepts inline bytes, data URLs, http(s) URLs${accepted}`))
+}
+
 /**
 * Lift a tool-result file into a `MediaPart`. Tool files carry either a data URL, an `http(s)` URL, or raw base64 in
 * `uri`; the declared `mime` wins over any data-URL prefix so tool authors control the type the model sees.
@@ -0,0 +1,113 @@
+import { Effect, Schema, type Stream } from "effect"
+import type { Media } from "../../media.js"
+import { Framing } from "../../route/framing.js"
+import type { MediaProtocol } from "../../route/media-protocol.js"
+import { AIError, ContentPolicyError, ProviderID, type MediaUsage, type ProviderMetadata } from "../../schema/index.js"
+import { ProviderShared } from "../shared.js"
+import { MediaInput } from "./media-input.js"
+
+const PROVIDER = ProviderID.make("google")
+
+const UsageMetadata = Schema.Struct({
+  promptTokenCount: Schema.optional(Schema.Number),
+  candidatesTokenCount: Schema.optional(Schema.Number),
+  totalTokenCount: Schema.optional(Schema.Number),
+})
+type UsageMetadata = Schema.Schema.Type<typeof UsageMetadata>
+
+export const chunk = <const Part extends Schema.Top>(part: Part) =>
+  Schema.Struct({
+    candidates: Schema.optional(
+      Schema.Array(
+        Schema.Struct({
+          content: Schema.optional(Schema.Struct({ parts: Schema.optional(Schema.Array(part)) })),
+          finishReason: Schema.optional(Schema.String),
+        }),
+      ),
+    ),
+    promptFeedback: Schema.optional(
+      Schema.Struct({
+        blockReason: Schema.optional(Schema.String),
+        blockReasonMessage: Schema.optional(Schema.String),
+      }),
+    ),
+    usageMetadata: Schema.optional(UsageMetadata),
+    modelVersion: Schema.optional(Schema.String),
+    responseId: Schema.optional(Schema.String),
+  })
+
+interface Chunk {
+  readonly candidates?: ReadonlyArray<{ readonly finishReason?: string }>
+  readonly promptFeedback?: { readonly blockReason?: string; readonly blockReasonMessage?: string }
+  readonly usageMetadata?: UsageMetadata
+  readonly modelVersion?: string
+  readonly responseId?: string
+}
+
+export interface Metadata {
+  readonly usage?: UsageMetadata
+  readonly finishReason?: string
+  readonly modelVersion?: string
+  readonly responseId?: string
+}
+
+export const track = <State extends Metadata>(state: State, chunk: Chunk): State => ({
+  ...state,
+  usage: chunk.usageMetadata ?? state.usage,
+  finishReason: chunk.candidates?.[0]?.finishReason ?? state.finishReason,
+  modelVersion: chunk.modelVersion ?? state.modelVersion,
+  responseId: chunk.responseId ?? state.responseId,
+})
+
+export const blocked = (name: string, chunk: Chunk, frame: string) => {
+  const feedback = chunk.promptFeedback
+  if (feedback?.blockReason === undefined) return undefined
+  return new AIError({
+    reason: new ContentPolicyError({
+      message: `${name} blocked the request (${feedback.blockReason})${
+        feedback.blockReasonMessage === undefined ? "" : `: ${feedback.blockReasonMessage}`
+      }`,
+      body: frame,
+    }),
+  })
+}
+
+export const usage = (usage: UsageMetadata | undefined): MediaUsage | undefined =>
+  usage === undefined
+    ? undefined
+    : {
+        type: "tokens",
+        input: usage.promptTokenCount,
+        output: usage.candidatesTokenCount,
+        total: ProviderShared.totalTokens(usage.promptTokenCount, usage.candidatesTokenCount, usage.totalTokenCount),
+        details: { google: usage },
+      }
+
+export const providerMetadata = (state: Metadata): ProviderMetadata => ({
+  google: { finishReason: state.finishReason, modelVersion: state.modelVersion, responseId: state.responseId },
+})
+
+export const path = (model: string, mode: MediaProtocol.Mode) =>
+  mode === "stream" ? `/models/${model}:streamGenerateContent?alt=sse` : `/models/${model}:generateContent`
+
+// `generateContent` answers with one document shaped exactly like a streamed chunk, so it is a single frame.
+export const frames = (bytes: Stream.Stream<Uint8Array, AIError>, mode: MediaProtocol.Mode) =>
+  mode === "stream" ? Framing.sse.frame(bytes) : Framing.document.frame(bytes)
+
+// Gemini does not fetch public URLs; inline payloads and Gemini Files references are the accepted inputs.
+export const mediaPart = (
+  route: string,
+  asset: Media.Asset,
+): Effect.Effect<
+  | { readonly fileData: { readonly mimeType: string; readonly fileUri: string } }
+  | { readonly inlineData: { readonly mimeType: string; readonly data: string } },
+  AIError
+> => {
+  const fileUri = MediaInput.refID(asset, PROVIDER)
+  if (fileUri !== undefined) return Effect.succeed({ fileData: { mimeType: asset.mediaType, fileUri } })
+  return ProviderShared.requireInlineMedia(route, asset).pipe(
+    Effect.map((media) => ({ inlineData: { mimeType: media.mime, data: media.base64 } })),
+  )
+}
+
+export * as GeminiGenerateContent from "./gemini-generate-content.js"
@@ -1,5 +1,6 @@
 import { Effect, Encoding } from "effect"
 import { Media } from "../../media.js"
+import type { MediaProtocol } from "../../route/media-protocol.js"
 import type { AIError, ProviderID } from "../../schema/index.js"
 import { ProviderShared } from "../shared.js"

@@ -13,6 +14,26 @@ export const inlineBytes = (route: string, asset: Media.Asset): Effect.Effect<Ui
  )
 }

+/** Copied because `BlobPart` requires a plain `ArrayBuffer`. */
+export const blob = (data: Uint8Array, mediaType: string) => {
+  const buffer = new ArrayBuffer(data.byteLength)
+  new Uint8Array(buffer).set(data)
+  return new Blob([buffer], { type: mediaType })
+}
+
+const isScalar = (value: unknown): value is string | number | boolean =>
+  typeof value === "string" || typeof value === "number" || typeof value === "boolean"
+
+export const query = (route: string, values: Record<string, unknown>): Effect.Effect<MediaProtocol.Query, AIError> => {
+  const entries = Object.entries(values).filter(([, value]) => value !== undefined)
+  const invalid = entries.find(([, value]) => !isScalar(value) && !(Array.isArray(value) && value.every(isScalar)))
+  if (invalid !== undefined)
+    return Effect.fail(ProviderShared.invalidRequest(`${route} cannot send "${invalid[0]}" as a query parameter`))
+  return Effect.succeed(
+    Object.fromEntries(entries.map(([key, value]) => [key, Array.isArray(value) ? value.map(String) : String(value)])),
+  )
+}
+
 /** Provider file handle when the ref belongs to this provider; refs from other providers are never forwarded. */
 export const refID = (asset: Media.Asset, provider: ProviderID) =>
  asset.source.type === "ref" && asset.source.provider === provider ? asset.source.id : undefined
@@ -0,0 +1,118 @@
+import { Effect } from "effect"
+import { Media } from "../../media.js"
+import { MediaProtocol } from "../../route/media-protocol.js"
+import type { AIError, MediaUsage, ProviderID, ProviderMetadata } from "../../schema/index.js"
+import {
+  SpeechAudioDeltaEvent,
+  SpeechFinishEvent,
+  SpeechTimestampsEvent,
+  type SpeechEvent,
+  type SpeechVoice,
+} from "../../speech.js"
+import { concatBytes } from "../../utils/bytes.js"
+import { ProviderShared } from "../shared.js"
+
+export interface Audio {
+  /** Appended in place: the route creates fresh state for each response through `initial`. */
+  readonly chunks: Array<Uint8Array>
+}
+
+export type StepResult<State> = readonly [State, ReadonlyArray<SpeechEvent>]
+
+/** Empty chunks (keep-alive records) emit nothing. */
+export const delta = <State extends Audio>(state: State, chunk: Uint8Array): StepResult<State> => {
+  if (chunk.length === 0) return [state, []]
+  state.chunks.push(chunk)
+  return [state, [SpeechAudioDeltaEvent.make({ chunk })]]
+}
+
+export const step =
+  <State extends Audio>(onRecord: (state: State, frame: string) => Effect.Effect<StepResult<State>, AIError>) =>
+  (state: State, frame: string | Uint8Array) =>
+    typeof frame === "string" ? onRecord(state, frame) : Effect.succeed(delta(state, frame))
+
+export const timestamps = (
+  texts: ReadonlyArray<string>,
+  starts: ReadonlyArray<number>,
+  ends: ReadonlyArray<number>,
+): ReadonlyArray<SpeechEvent> =>
+  texts.length === 0
+    ? []
+    : [
+        SpeechTimestampsEvent.make({
+          items: texts.map((text, index) => ({ text, startSeconds: starts[index] ?? 0, endSeconds: ends[index] ?? 0 })),
+        }),
+      ]
+
+export const voiceID = (voice: SpeechVoice | undefined) => (typeof voice === "object" ? voice.id : voice)
+
+const CONTAINER_MEDIA_TYPES: Readonly<Record<string, string>> = {
+  mp3: "audio/mpeg",
+  wav: "audio/wav",
+  opus: "audio/ogg",
+  aac: "audio/aac",
+  flac: "audio/flac",
+}
+
+export const container = (format: string, sampleRate?: number) => ({
+  mediaType: CONTAINER_MEDIA_TYPES[format],
+  info: { format, sampleRate },
+})
+
+const PCM_MEDIA_TYPES = {
+  pcm_s16le: "audio/pcm",
+  pcm_f32le: "audio/pcm",
+  pcm_mulaw: "audio/mulaw",
+  pcm_alaw: "audio/alaw",
+} as const
+
+export type PcmEncoding = keyof typeof PCM_MEDIA_TYPES
+
+export const pcm = (encoding: PcmEncoding, sampleRate: number | undefined, mediaType?: string) => ({
+  mediaType: mediaType ?? PCM_MEDIA_TYPES[encoding],
+  info: { format: "pcm", encoding, sampleRate, channels: 1 },
+})
+
+export const sampleRate = (mediaType: string | undefined) => {
+  const rate = /rate=(\d+)/i.exec(mediaType ?? "")?.[1]
+  return rate === undefined ? undefined : Number(rate)
+}
+
+export const unsupportedFormat = (provider: ProviderID, route: string, message: string) =>
+  ProviderShared.unsupportedOperation({ operation: "media.format", provider, route, message })
+
+/** A declared `mediaType` wins over sniffing: headerless PCM can start with bytes that look like an MPEG frame sync. */
+export const finish = (
+  route: string,
+  state: Audio,
+  output: {
+    readonly mediaType: string | undefined
+    readonly info?: Media.Info
+    readonly usage?: MediaUsage
+    readonly providerMetadata?: ProviderMetadata
+    readonly detail?: string
+  },
+): Effect.Effect<ReadonlyArray<SpeechEvent>, AIError> => {
+  if (state.chunks.length === 0)
+    return Effect.fail(
+      MediaProtocol.frameError(
+        route,
+        `The provider returned no audio${output.detail === undefined ? "" : ` (${output.detail})`}`,
+      ),
+    )
+  return Effect.succeed([
+    SpeechFinishEvent.make({
+      audio: Media.bytes(concatBytes(state.chunks), output.mediaType, { info: output.info }),
+      usage: output.usage,
+      providerMetadata: output.providerMetadata,
+    }),
+  ])
+}
+
+export const headerUsage = (type: "characters" | "credits", value: string | undefined): MediaUsage | undefined => {
+  const amount = Number(value)
+  if (!Number.isFinite(amount)) return undefined
+  return type === "credits" ? { type, credits: amount } : { type, characters: amount }
+}
+
+export * as SpeechStream from "./speech-stream.js"
@@ -4,7 +4,7 @@ import { ImageModel, ImageResponse, type ImageRequestFor } from "../image.js"
 import { Media } from "../media.js"
 import { MediaProtocol } from "../route/media-protocol.js"
 import { MediaRoute } from "../route/media.js"
-import { ProviderID, mergeJsonRecords, type AIError } from "../schema/index.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
 import { ProviderShared, optionalNull } from "./shared.js"
 import { MediaInput } from "./utils/media-input.js"

@@ -58,15 +58,12 @@ const nativeOptions = (options: XAIImageOptions | undefined) => {

 const isEdit = (request: Request) => (request.images?.length ?? 0) > 0

-const reference = (asset: Media.Asset): Effect.Effect<Record<string, unknown>, AIError> => {
-  const inline = asset.inline()
-  if (inline) return Effect.succeed({ url: inline.dataUrl, type: "image_url" as const })
-  const url = ProviderShared.mediaUrl(asset)
-  if (url) return Effect.succeed({ url, type: "image_url" as const })
-  const id = MediaInput.refID(asset, PROVIDER)
-  if (id) return Effect.succeed({ file_id: id })
-  return Effect.fail(ProviderShared.invalidRequest(`${NAME} accepts image URLs, data URLs, bytes, and xAI file IDs`))
-}
+const reference = (asset: Media.Asset) =>
+  ProviderShared.mediaReference(asset, PROVIDER, NAME).pipe(
+    Effect.map((item) =>
+      item.type === "ref" ? { file_id: item.value } : { url: item.value, type: "image_url" as const },
+    ),
+  )

 const fromRequest = Effect.fn("XAIImages.fromRequest")(function* (request: Request) {
  const references = yield* Effect.forEach(request.images ?? [], reference)
@@ -0,0 +1,211 @@
+import { Effect, Schema } from "effect"
+import type { HttpClientResponse } from "effect/unstable/http"
+import type { Status } from "../generation.js"
+import { Media } from "../media.js"
+import { MediaProtocol } from "../route/media-protocol.js"
+import { MediaRoute } from "../route/media.js"
+import { ProviderID, mergeJsonRecords } from "../schema/index.js"
+import { VideoModel, VideoResponse, type VideoRequestFor } from "../video.js"
+import { ProviderShared, optionalNull } from "./shared.js"
+
+const ADAPTER = "xai-video"
+const NAME = "xAI Video"
+const PROVIDER = ProviderID.make("xai")
+export const DEFAULT_BASE_URL = "https://api.x.ai/v1"
+export const PATH = "/videos/generations"
+export const EDIT_PATH = "/videos/edits"
+export const EXTEND_PATH = "/videos/extensions"
+export const STATUS_PATH = "/videos"
+
+// ---------------------------------------------------------------------------
+// 1. Public model input
+// ---------------------------------------------------------------------------
+
+/**
+ * Provider-native options. Common fields (`frames`, `references`, `video`, `durationSeconds`, `aspectRatio`,
+ * `resolution`, `audio`) live on the request. `mode` selects the endpoint a `video` input is sent to.
+ */
+export type XAIVideoOptions = {
+  readonly mode?: "edit" | "extend"
+  readonly reference_audios?: ReadonlyArray<{ readonly voice_id: string }>
+} & Record<string, unknown>
+
+export type Request = VideoRequestFor<XAIVideoOptions>
+
+// ---------------------------------------------------------------------------
+// 2. Token and response schemas
+// ---------------------------------------------------------------------------
+
+export const Token = Schema.Struct({ requestID: Schema.String })
+export type Token = Schema.Schema.Type<typeof Token>
+
+const StartResponse = Schema.Struct({ request_id: Schema.String })
+
+const VideoStatus = Schema.Struct({
+  status: Schema.String,
+  progress: optionalNull(Schema.Number),
+  video: optionalNull(
+    Schema.Struct({
+      url: optionalNull(Schema.String),
+      duration: optionalNull(Schema.Number),
+      respect_moderation: optionalNull(Schema.Boolean),
+    }),
+  ),
+  error: optionalNull(
+    Schema.Struct({
+      code: optionalNull(Schema.String),
+      message: optionalNull(Schema.String),
+    }),
+  ),
+  model: optionalNull(Schema.String),
+})
+
+const STATUS = {
+  pending: "running",
+  done: "completed",
+  failed: "failed",
+  expired: "expired",
+} as const satisfies Record<string, Status>
+
+// ---------------------------------------------------------------------------
+// 5. Request body construction
+// ---------------------------------------------------------------------------
+
+const mediaInput = (asset: Media.Asset) =>
+  ProviderShared.mediaReference(asset, PROVIDER, NAME).pipe(
+    Effect.map((reference) => (reference.type === "ref" ? { file_id: reference.value } : { url: reference.value })),
+  )
+
+const nativeOptions = (options: XAIVideoOptions | undefined) => {
+  if (!options) return undefined
+  const { mode: _mode, ...native } = options
+  return native
+}
+
+const fromRequest = Effect.fn("XAIVideo.fromRequest")(function* (request: Request) {
+  const image = request.frames?.first === undefined ? undefined : yield* mediaInput(request.frames.first)
+  const lastFrame = request.frames?.last === undefined ? undefined : yield* mediaInput(request.frames.last)
+  const video = request.video === undefined ? undefined : yield* mediaInput(request.video)
+  const references = yield* Effect.forEach(request.references ?? [], mediaInput)
+  return MediaProtocol.json(
+    mergeJsonRecords(
+      {
+        model: request.model.id,
+        prompt: request.prompt,
+        image,
+        last_frame: lastFrame,
+        reference_images: references.length === 0 ? undefined : references,
+        video,
+        duration: request.durationSeconds,
+        aspect_ratio: request.aspectRatio,
+        resolution: request.resolution,
+        generate_audio: request.audio,
+      },
+      nativeOptions(request.providerOptions),
+      request.http?.body,
+    ) ?? {},
+  )
+})
+
+// ---------------------------------------------------------------------------
+// 6. Response decoding
+// ---------------------------------------------------------------------------
+
+const decodeStart = MediaProtocol.decodeStarted(ADAPTER, NAME, StartResponse, (value) => ({
+  token: { requestID: value.request_id },
+  snapshot: { id: value.request_id, status: "running" },
+}))
+
+// `progress` is undocumented but observed live as a 0..100 percentage (recorded cassette: 1 → 10 → 37 → 100).
+const fraction = (progress: number | null | undefined) =>
+  progress !== undefined && progress !== null && progress >= 0 && progress <= 100 ? progress / 100 : undefined
+
+const decodeVideoStatus = MediaProtocol.decodeJson(ADAPTER, NAME, VideoStatus)
+
+const decodeStatus = Effect.fn("XAIVideo.decodeStatus")(function* (
+  response: HttpClientResponse.HttpClientResponse,
+  context: MediaProtocol.PollContext<Token>,
+) {
+  const output = yield* decodeVideoStatus(response)
+  const status = yield* MediaProtocol.status(STATUS, output.value.status, output)
+  return { id: context.token.requestID, status, progress: fraction(output.value.progress) }
+})
+
+const decodeResult = Effect.fn("XAIVideo.decodeResult")(function* (
+  response: HttpClientResponse.HttpClientResponse,
+  context: MediaProtocol.PollContext<Token>,
+) {
+  const output = yield* decodeVideoStatus(response)
+  const decoded = output.value
+  const status = yield* MediaProtocol.status(STATUS, decoded.status, output)
+  if (status === "running") return yield* output.invalid(`${NAME} request ${context.token.requestID} has not finished`)
+  if (status === "failed") {
+    const code = decoded.error?.code ?? undefined
+    const message = decoded.error?.message ?? undefined
+    return yield* output.ended(
+      "failed",
+      `${NAME} generation failed${code === undefined ? "" : ` (${code})`}${message === undefined ? "" : `: ${message}`}`,
+    )
+  }
+  if (status !== "completed")
+    return yield* output.ended("expired", `${NAME} request ${context.token.requestID} expired`)
+  // `respect_moderation: false` marks a filtered result; a URL may still be present, so report it as a notice.
+  const notices =
+    decoded.video?.respect_moderation === false
+      ? [{ type: "moderated" as const, message: `${NAME} flagged the generated video for moderation` }]
+      : undefined
+  const url = decoded.video?.url ?? undefined
+  if (url === undefined && notices !== undefined)
+    return yield* output.contentPolicy(`${NAME} withheld the video for moderation`)
+  if (url === undefined) return yield* output.invalid(`${NAME} completed without a video URL`)
+  const duration = decoded.video?.duration ?? undefined
+  return new VideoResponse({
+    videos: [
+      Media.url(url, {
+        mediaType: "video/mp4",
+        info: duration === undefined ? undefined : { durationSeconds: duration },
+      }),
+    ],
+    notices,
+    providerMetadata: { xai: { requestId: context.token.requestID, model: decoded.model ?? undefined } },
+  })
+})
+
+// ---------------------------------------------------------------------------
+// 7. Protocol and route
+// ---------------------------------------------------------------------------
+
+const statusPath = (token: Token) => `${STATUS_PATH}/${token.requestID}`
+
+export const protocol = MediaProtocol.queued<Request, VideoResponse, Token>({
+  id: ADAPTER,
+  name: NAME,
+  token: Token,
+  unsupported: ["n", "seed", "negativePrompt"],
+  start: { body: { from: fromRequest }, decode: decodeStart },
+  status: { path: statusPath, decode: decodeStatus },
+  result: { path: statusPath, decode: decodeResult },
+})
+
+// A source video goes to `/videos/edits` unless `providerOptions.mode` asks for an extension.
+const startPath = (request: Request) => {
+  if (request.video === undefined) return PATH
+  return request.providerOptions?.mode === "extend" ? EXTEND_PATH : EDIT_PATH
+}
+
+export const model = (input: MediaRoute.ModelInput) =>
+  VideoModel.fromRoute<XAIVideoOptions, Token>(
+    {
+      id: ADAPTER,
+      provider: PROVIDER,
+      protocol,
+      baseURL: DEFAULT_BASE_URL,
+      path: ({ request }) => startPath(request),
+    },
+    input,
+  )
+
+export const XAIVideo = {
+  protocol,
+  model,
+} as const
@@ -0,0 +1,43 @@
+import { Auth } from "../route/auth.js"
+import type { ProviderAuthOption } from "../route/auth-options.js"
+import { HttpOptions, ProviderID, type ModelID } from "../schema/index.js"
+import { AssemblyAITranscription, DEFAULT_BASE_URL } from "../protocols/assemblyai-transcription.js"
+
+export type { AssemblyAITranscriptionOptions } from "../protocols/assemblyai-transcription.js"
+
+export const id = ProviderID.make("assemblyai")
+const baseURL = DEFAULT_BASE_URL
+
+export type Config = ProviderAuthOption<"optional"> & {
+  /** `https://api.eu.assemblyai.com` for the EU region. */
+  readonly baseURL?: string
+  readonly headers?: Record<string, string>
+  readonly http?: HttpOptions.Input
+}
+
+// The key is the whole `authorization` value, without a scheme.
+const auth = (options: ProviderAuthOption<"optional">) => {
+  if ("auth" in options && options.auth) return options.auth
+  return Auth.optional("apiKey" in options ? options.apiKey : undefined, "apiKey")
+    .orElse(Auth.config("ASSEMBLYAI_API_KEY"))
+    .pipe(Auth.header("authorization"))
+}
+
+export const configure = (input: Config = {}) => {
+  const transcription = (modelID: string | ModelID) =>
+    AssemblyAITranscription.model({
+      id: modelID,
+      auth: auth(input),
+      baseURL: input.baseURL ?? baseURL,
+      headers: input.headers,
+      http: input.http === undefined ? undefined : HttpOptions.make(input.http),
+    })
+  return {
+    id,
+    transcription,
+    configure,
+  }
+}
+
+export const provider = configure()
+export const transcription = provider.transcription
@@ -0,0 +1,35 @@
+import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
+import { HttpOptions, ProviderID, type ModelID } from "../schema/index.js"
+import { CartesiaSpeech, DEFAULT_BASE_URL } from "../protocols/cartesia-speech.js"
+
+export type { CartesiaEncoding, CartesiaSpeechOptions } from "../protocols/cartesia-speech.js"
+
+export const id = ProviderID.make("cartesia")
+const baseURL = DEFAULT_BASE_URL
+
+export type Config = ProviderAuthOption<"optional"> & {
+  readonly baseURL?: string
+  readonly headers?: Record<string, string>
+  readonly http?: HttpOptions.Input
+}
+
+const auth = (options: ProviderAuthOption<"optional">) => AuthOptions.bearer(options, "CARTESIA_API_KEY")
+
+export const configure = (input: Config = {}) => {
+  const speech = (modelID: string | ModelID) =>
+    CartesiaSpeech.model({
+      id: modelID,
+      auth: auth(input),
+      baseURL: input.baseURL ?? baseURL,
+      headers: input.headers,
+      http: input.http === undefined ? undefined : HttpOptions.make(input.http),
+    })
+  return {
+    id,
+    speech,
+    configure,
+  }
+}
+
+export const provider = configure()
+export const speech = provider.speech
@@ -0,0 +1,44 @@
+import { Auth } from "../route/auth.js"
+import type { ProviderAuthOption } from "../route/auth-options.js"
+import { HttpOptions, ProviderID, type ModelID } from "../schema/index.js"
+import { DEFAULT_BASE_URL, DeepgramSpeech } from "../protocols/deepgram-speech.js"
+import { DeepgramTranscription } from "../protocols/deepgram-transcription.js"
+
+export type { DeepgramEncoding, DeepgramSpeechOptions } from "../protocols/deepgram-speech.js"
+export type { DeepgramTranscriptionOptions } from "../protocols/deepgram-transcription.js"
+
+export const id = ProviderID.make("deepgram")
+const baseURL = DEFAULT_BASE_URL
+
+export type Config = ProviderAuthOption<"optional"> & {
+  readonly baseURL?: string
+  readonly headers?: Record<string, string>
+  readonly http?: HttpOptions.Input
+}
+
+const auth = (options: ProviderAuthOption<"optional">) => {
+  if ("auth" in options && options.auth) return options.auth
+  return Auth.optional("apiKey" in options ? options.apiKey : undefined, "apiKey")
+    .orElse(Auth.config("DEEPGRAM_API_KEY"))
+    .pipe(Auth.scheme("Token"))
+}
+
+export const configure = (input: Config = {}) => {
+  const media = (modelID: string | ModelID) => ({
+    id: modelID,
+    auth: auth(input),
+    baseURL: input.baseURL ?? baseURL,
+    headers: input.headers,
+    http: input.http === undefined ? undefined : HttpOptions.make(input.http),
+  })
+  return {
+    id,
+    speech: (modelID: string | ModelID) => DeepgramSpeech.model(media(modelID)),
+    transcription: (modelID: string | ModelID) => DeepgramTranscription.model(media(modelID)),
+    configure,
+  }
+}
+
+export const provider = configure()
+export const speech = provider.speech
+export const transcription = provider.transcription
@@ -0,0 +1,41 @@
+import { Auth } from "../route/auth.js"
+import type { ProviderAuthOption } from "../route/auth-options.js"
+import { HttpOptions, ProviderID, type ModelID } from "../schema/index.js"
+import { DEFAULT_BASE_URL, ElevenLabsSpeech } from "../protocols/elevenlabs-speech.js"
+
+export type { ElevenLabsOutputFormat, ElevenLabsSpeechOptions } from "../protocols/elevenlabs-speech.js"
+
+export const id = ProviderID.make("elevenlabs")
+const baseURL = DEFAULT_BASE_URL
+
+export type Config = ProviderAuthOption<"optional"> & {
+  readonly baseURL?: string
+  readonly headers?: Record<string, string>
+  readonly http?: HttpOptions.Input
+}
+
+const auth = (options: ProviderAuthOption<"optional">) => {
+  if ("auth" in options && options.auth) return options.auth
+  return Auth.optional("apiKey" in options ? options.apiKey : undefined, "apiKey")
+    .orElse(Auth.config("ELEVENLABS_API_KEY"))
+    .pipe(Auth.header("xi-api-key"))
+}
+
+export const configure = (input: Config = {}) => {
+  const speech = (modelID: string | ModelID) =>
+    ElevenLabsSpeech.model({
+      id: modelID,
+      auth: auth(input),
+      baseURL: input.baseURL ?? baseURL,
+      headers: input.headers,
+      http: input.http === undefined ? undefined : HttpOptions.make(input.http),
+    })
+  return {
+    id,
+    speech,
+    configure,
+  }
+}
+
+export const provider = configure()
+export const speech = provider.speech
@@ -0,0 +1,42 @@
+import { Auth } from "../route/auth.js"
+import type { ProviderAuthOption } from "../route/auth-options.js"
+import { HttpOptions, ProviderID, type ModelID } from "../schema/index.js"
+import { DEFAULT_BASE_URL, FalVideo } from "../protocols/fal-video.js"
+
+export type { FalVideoOptions } from "../protocols/fal-video.js"
+
+export const id = ProviderID.make("fal")
+const baseURL = DEFAULT_BASE_URL
+
+export type Config = ProviderAuthOption<"optional"> & {
+  readonly baseURL?: string
+  readonly headers?: Record<string, string>
+  readonly http?: HttpOptions.Input
+}
+
+// fal authenticates with `Authorization: Key <FAL_KEY>` rather than a bearer token.
+const auth = (options: ProviderAuthOption<"optional">) => {
+  if ("auth" in options && options.auth) return options.auth
+  return Auth.optional("apiKey" in options ? options.apiKey : undefined, "apiKey")
+    .orElse(Auth.config("FAL_KEY"))
+    .pipe(Auth.scheme("Key"))
+}
+
+export const configure = (input: Config = {}) => {
+  const video = (modelID: string | ModelID) =>
+    FalVideo.model({
+      id: modelID,
+      auth: auth(input),
+      baseURL: input.baseURL ?? baseURL,
+      headers: input.headers,
+      http: input.http === undefined ? undefined : HttpOptions.make(input.http),
+    })
+  return {
+    id,
+    video,
+    configure,
+  }
+}
+
+export const provider = configure()
+export const video = provider.video
@@ -5,8 +5,14 @@ import type { ProviderPackage } from "../provider-package.js"
 import { HttpOptions, ProviderID, mergeHttpOptions, type ModelID } from "../schema/index.js"
 import { Gemini } from "../protocols/gemini.js"
 import { GoogleImages } from "../protocols/google-images.js"
+import { GoogleSpeech } from "../protocols/google-speech.js"
+import { GoogleTranscription } from "../protocols/google-transcription.js"
+import { GoogleVideo } from "../protocols/google-video.js"

 export type { GoogleImageOptions } from "../protocols/google-images.js"
+export type { GoogleSpeechOptions } from "../protocols/google-speech.js"
+export type { GoogleTranscriptionOptions } from "../protocols/google-transcription.js"
+export type { GoogleVideoOptions } from "../protocols/google-video.js"
 export type GeminiOptionsInput = Gemini.OptionsInput
 export type GeminiProviderOptionsInput = Gemini.ProviderOptionsInput

@@ -40,18 +46,20 @@ const configuredRoute = (input: Config) => {

 export const configure = (input: Config = {}) => {
  const route = configuredRoute(input)
-  const image = (modelID: string | ModelID) =>
-    GoogleImages.model({
-      id: modelID,
-      auth: auth(input),
-      baseURL: input.baseURL,
-      headers: input.headers,
-      http: mergeHttpOptions(input.http === undefined ? undefined : HttpOptions.make(input.http)),
-    })
+  const media = (modelID: string | ModelID) => ({
+    id: modelID,
+    auth: auth(input),
+    baseURL: input.baseURL,
+    headers: input.headers,
+    http: mergeHttpOptions(input.http === undefined ? undefined : HttpOptions.make(input.http)),
+  })
  return {
    id,
    model: (modelID: string | ModelID) => route.model<Gemini.ProviderOptionsInput>({ id: modelID }),
-    image,
+    image: (modelID: string | ModelID) => GoogleImages.model(media(modelID)),
+    video: (modelID: string | ModelID) => GoogleVideo.model(media(modelID)),
+    speech: (modelID: string | ModelID) => GoogleSpeech.model(media(modelID)),
+    transcription: (modelID: string | ModelID) => GoogleTranscription.model(media(modelID)),
    configure,
  }
 }
@@ -70,3 +78,6 @@ export const model: ProviderPackage.Definition<Settings, Gemini.ProviderOptionsI
  }).model(modelID)

 export const image = provider.image
+export const video = provider.video
+export const speech = provider.speech
+export const transcription = provider.transcription
@@ -3,13 +3,18 @@ export * as Anthropic from "./anthropic.js"
 export * as AnthropicCompatible from "./anthropic-compatible.js"
 export * as AmazonBedrock from "./amazon-bedrock.js"
 export * as AmazonBedrockMantle from "./amazon-bedrock-mantle.js"
+export * as AssemblyAI from "./assemblyai.js"
 export * as Azure from "./azure.js"
 export * as Baseten from "./baseten.js"
+export * as Cartesia from "./cartesia.js"
 export * as Cerebras from "./cerebras.js"
 export * as CloudflareAIGateway from "./cloudflare-ai-gateway.js"
 export * as CloudflareWorkersAI from "./cloudflare-workers-ai.js"
 export * as DeepInfra from "./deepinfra.js"
+export * as Deepgram from "./deepgram.js"
 export * as DeepSeek from "./deepseek.js"
+export * as ElevenLabs from "./elevenlabs.js"
+export * as Fal from "./fal.js"
 export * as Fireworks from "./fireworks.js"
 export * as Google from "./google.js"
 export * as GoogleVertex from "./google-vertex.js"
@@ -26,6 +31,7 @@ export * as OpenAICompatible from "./openai-compatible.js"
 export * as OpenAICompatibleResponses from "./openai-compatible-responses.js"
 export * as OpenCodeZen from "./opencode-zen.js"
 export * as OpenRouter from "./openrouter.js"
+export * as Runway from "./runway.js"
 export * as TogetherAI from "./togetherai.js"
 export * as TypeSafeAI from "./typesafe-ai.js"
 export * as VercelAIGateway from "./vercel-ai-gateway.js"
@@ -6,9 +6,13 @@ import * as OpenAIChat from "../protocols/openai-chat.js"
 import * as OpenAIResponses from "../protocols/openai-responses.js"
 import { withOpenAIOptions, type OpenAIProviderOptionsInput } from "./openai-options.js"
 import { OpenAIImages, type OpenAIImageString } from "../protocols/openai-images.js"
+import { OpenAISpeech } from "../protocols/openai-speech.js"
+import { OpenAITranscription } from "../protocols/openai-transcription.js"

 export type { OpenAIOptionsInput, OpenAIResponseIncludable } from "./openai-options.js"
 export type { OpenAIImageOptions } from "../protocols/openai-images.js"
+export type { OpenAISpeechOptions } from "../protocols/openai-speech.js"
+export type { OpenAITranscriptionOptions } from "../protocols/openai-transcription.js"

 export const id = ProviderID.make("openai")

@@ -95,17 +99,19 @@ export const configure = (input: Config = {}) => {
      id,
      compatibility: { supportsPromptCacheKey: true },
    })
-  const image = (modelID: string | ModelID) =>
-    OpenAIImages.model({
-      id: modelID,
-      auth: auth(input),
-      baseURL: input.baseURL,
-      headers: input.headers,
-      http: mergeHttpOptions(
-        input.http === undefined ? undefined : HttpOptions.make(input.http),
-        input.queryParams === undefined ? undefined : new HttpOptions({ query: input.queryParams }),
-      ),
-    })
+  const media = (modelID: string | ModelID) => ({
+    id: modelID,
+    auth: auth(input),
+    baseURL: input.baseURL,
+    headers: input.headers,
+    http: mergeHttpOptions(
+      input.http === undefined ? undefined : HttpOptions.make(input.http),
+      input.queryParams === undefined ? undefined : new HttpOptions({ query: input.queryParams }),
+    ),
+  })
+  const image = (modelID: string | ModelID) => OpenAIImages.model(media(modelID))
+  const speech = (modelID: string | ModelID) => OpenAISpeech.model(media(modelID))
+  const transcription = (modelID: string | ModelID) => OpenAITranscription.model(media(modelID))

  return {
    id,
@@ -113,6 +119,8 @@ export const configure = (input: Config = {}) => {
    responses,
    chat,
    image,
+    speech,
+    transcription,
    configure,
  }
 }
@@ -159,3 +167,5 @@ export const chatModel: ProviderPackage.Definition<Settings, OpenAIProviderOptio
 export const responses = provider.responses
 export const chat = provider.chat
 export const image = provider.image
+export const speech = provider.speech
+export const transcription = provider.transcription
@@ -0,0 +1,35 @@
+import { AuthOptions, type ProviderAuthOption } from "../route/auth-options.js"
+import { HttpOptions, ProviderID, type ModelID } from "../schema/index.js"
+import { DEFAULT_BASE_URL, RunwayVideo } from "../protocols/runway-video.js"
+
+export type { RunwayVideoOptions } from "../protocols/runway-video.js"
+
+export const id = ProviderID.make("runway")
+const baseURL = DEFAULT_BASE_URL
+
+export type Config = ProviderAuthOption<"optional"> & {
+  readonly baseURL?: string
+  readonly headers?: Record<string, string>
+  readonly http?: HttpOptions.Input
+}
+
+const auth = (options: ProviderAuthOption<"optional">) => AuthOptions.bearer(options, "RUNWAYML_API_SECRET")
+
+export const configure = (input: Config = {}) => {
+  const video = (modelID: string | ModelID) =>
+    RunwayVideo.model({
+      id: modelID,
+      auth: auth(input),
+      baseURL: input.baseURL ?? baseURL,
+      headers: input.headers,
+      http: input.http === undefined ? undefined : HttpOptions.make(input.http),
+    })
+  return {
+    id,
+    video,
+    configure,
+  }
+}
+
+export const provider = configure()
+export const video = provider.video
@@ -6,6 +6,7 @@ import { OpenAIChat } from "../protocols/openai-chat.js"
 import { OpenResponsesChannel } from "../protocols/open-responses-channel.js"
 import { XAIResponses } from "../protocols/xai-responses.js"
 import { XAIImages } from "../protocols/xai-images.js"
+import { XAIVideo } from "../protocols/xai-video.js"
 import type { OpenAIOptionsInput } from "./openai-options.js"
 import type { ProviderPackage } from "../provider-package.js"

@@ -27,6 +28,7 @@ export type Settings = ProviderPackage.Settings &
  }

 export type { XAIImageOptions } from "../protocols/xai-images.js"
+export type { XAIVideoOptions } from "../protocols/xai-video.js"

 const RESPONSES_WEBSOCKET_ROTATE_AFTER_MS = 24 * 60 * 1000

@@ -87,20 +89,20 @@ export const configure = (input: LanguageModelOptions = {}) => {
  const chatRoute = configuredChatRoute(input)
  const responses = (modelID: string | ModelID) => responsesRoute.model<XAIProviderOptionsInput>({ id: modelID })
  const chat = (modelID: string | ModelID) => chatRoute.model<XAIProviderOptionsInput>({ id: modelID })
-  const image = (modelID: string | ModelID) =>
-    XAIImages.model({
-      id: modelID,
-      auth: auth(input),
-      baseURL: input.baseURL ?? baseURL,
-      headers: input.headers,
-      http: input.http === undefined ? undefined : HttpOptions.make(input.http),
-    })
+  const media = (modelID: string | ModelID) => ({
+    id: modelID,
+    auth: auth(input),
+    baseURL: input.baseURL ?? baseURL,
+    headers: input.headers,
+    http: input.http === undefined ? undefined : HttpOptions.make(input.http),
+  })
  return {
    id,
    model: responses,
    responses,
    chat,
-    image,
+    image: (modelID: string | ModelID) => XAIImages.model(media(modelID)),
+    video: (modelID: string | ModelID) => XAIVideo.model(media(modelID)),
    configure,
  }
 }
@@ -121,3 +123,4 @@ export const model: ProviderPackage.Definition<
 export const responses = provider.responses
 export const chat = provider.chat
 export const image = provider.image
+export const video = provider.video
@@ -16,7 +16,7 @@ type Secret = string | Redacted.Redacted | Config.Config<string | Redacted.Redac

 export interface AuthInput {
  readonly request: { readonly http?: HttpOptions }
-  readonly method: "POST" | "GET"
+  readonly method: "POST" | "GET" | "PUT" | "DELETE"
  readonly url: string
  readonly body: string
  readonly headers: Headers.Headers
@@ -134,6 +134,16 @@ export function bearerHeader(name: string, source?: Secret | Credential) {
  return render(source)
 }

+/** `Authorization: <scheme> <secret>` for providers whose scheme is not `Bearer`, such as fal's `Key`. */
+export function scheme(name: string): (source: Secret | Credential) => Definition
+export function scheme(name: string, source: Secret | Credential): Definition
+export function scheme(name: string, source?: Secret | Credential) {
+  const render = (input: Secret | Credential) =>
+    fromCredential(credentialInput(input), (secret) => ({ authorization: `${name} ${secret}` }))
+  if (source === undefined) return render
+  return render(source)
+}
+
 const toAIError = (error: AuthError): AIError => {
  if (error instanceof MissingCredentialError || error instanceof Config.ConfigError) {
    return new AIError({
@@ -1,4 +1,4 @@
-import type { Stream } from "effect"
+import { Stream } from "effect"
 import * as ProviderShared from "../protocols/shared.js"
 import type { AIError } from "../schema/index.js"

@@ -12,6 +12,8 @@ import type { AIError } from "../schema/index.js"
 *   `[DONE]`; protocols that use it as a terminal select `sseWithDone`.
 * - AWS event stream — length-prefixed binary frames with CRC checksums.
 *   Each emitted frame is one parsed binary event record.
+ * - Media streams — newline-delimited JSON (`lines`) or the whole body as one
+ *   frame (`document`); chunked binary bodies need no framing.
 *
 * The frame type is opaque to this layer; the protocol's event schema decodes
 * each frame before its state machine handles it.
@@ -38,4 +40,19 @@ export const sseEvents = (events: ReadonlySet<string>): Definition<string> => ({
  frame: (bytes) => ProviderShared.sseFraming(bytes, events),
 })

+export const lines: Definition<string> = {
+  id: "lines",
+  frame: (bytes) =>
+    bytes.pipe(
+      Stream.decodeText(),
+      Stream.splitLines,
+      Stream.filter((line) => line.trim().length > 0),
+    ),
+}
+
+export const document: Definition<string> = {
+  id: "document",
+  frame: (bytes) => Stream.fromEffect(Stream.mkString(bytes.pipe(Stream.decodeText()))),
+}
+
 export * as Framing from "./framing.js"
@@ -1,17 +1,52 @@
-import { Effect, Schema } from "effect"
+import { Clock, Duration, Effect, Schema, type Stream } from "effect"
 import { HttpClientResponse } from "effect/unstable/http"
-import { AIError, HttpContext, InvalidProviderOutputError } from "../schema/index.js"
+import type { Snapshot, Status } from "../generation.js"
+import { Media } from "../media.js"
+import type { AuthInput } from "./auth.js"
+import {
+  AIError,
+  ContentPolicyError,
+  HttpContext,
+  InvalidProviderOutputError,
+  InvalidRequestError,
+  ProviderInternalError,
+} from "../schema/index.js"

 // ---------------------------------------------------------------------------
 // Bodies
 // ---------------------------------------------------------------------------

-export type Body =
-  | { readonly type: "json"; readonly value: Record<string, unknown> }
-  | { readonly type: "multipart"; readonly value: FormData }
+/** Array values become repeated parameters (`keyterm=a&keyterm=b`). */
+export type Query = Readonly<Record<string, string | ReadonlyArray<string>>>

-export const json = (value: Record<string, unknown>): Body => ({ type: "json", value })
+/** `query` is appended to the endpoint URL before the route and caller `http.query` overlays. */
+export type Body =
+  | { readonly type: "json"; readonly value: Record<string, unknown>; readonly query?: Query }
+  | { readonly type: "multipart"; readonly value: FormData }
+  | {
+      readonly type: "binary"
+      readonly value: Uint8Array
+      readonly contentType: string
+      readonly query?: Query
+    }
+
+export const json = (value: Record<string, unknown>, query?: Query): Body => ({
+  type: "json",
+  value,
+  query,
+})
 export const multipart = (value: FormData): Body => ({ type: "multipart", value })
+export const binary = (value: Uint8Array, contentType: string, query?: Query): Body => ({
+  type: "binary",
+  value,
+  contentType,
+  query,
+})
+
+export type Send = (path: string, body: Body) => Effect.Effect<HttpClientResponse.HttpClientResponse, AIError>
+
+/** Runs after unsupported-field rejection and before `body.from`, for providers that need an upload first. */
+export type Prepare<Request> = (request: Request, send: Send) => Effect.Effect<Request, AIError>

 // ---------------------------------------------------------------------------
 // Protocol kinds
@@ -45,6 +80,107 @@ export const inline = <Request, Response>(
  ...input,
 })

+/** What `start` learned from the submission response: the route-owned handle plus the first observation. */
+export interface Started<Token> {
+  readonly token: Token
+  readonly snapshot: Snapshot
+}
+
+/**
+ * A follow-up call's inputs: the decoded token and the auth headers the route sent, so a protocol can attach them
+ * to output URLs that require the same credentials to download (Veo).
+ */
+export interface PollContext<Token> {
+  readonly token: Token
+  readonly auth: Record<string, string>
+}
+
+/**
+ * Submit, then poll. `start` posts the body to the route endpoint; `status`, `result`, and `cancel` are follow-up
+ * calls addressed by the token. Paths are relative to the route base URL unless the provider hands back absolute
+ * URLs (fal `status_url`), in which case they are used verbatim. `result` is always its own GET: providers that
+ * return the output inside the status body (Veo, xAI, Runway) point `result.path` at the status path and decode the
+ * same document, so `Generation.await` and `Video.resume(...).await()` behave identically everywhere.
+ */
+export interface Queued<Request, Response, Token> {
+  readonly kind: "queued"
+  readonly id: string
+  readonly name: string
+  /** Common request fields this protocol cannot lower; the route rejects them before `start.body.from` runs. */
+  readonly unsupported?: ReadonlyArray<keyof Request & string>
+  /** Serializable handle. `Generation.token` carries the encoded form so it can be persisted and resumed elsewhere. */
+  readonly token: Schema.Codec<Token, unknown>
+  readonly start: {
+    readonly prepare?: Prepare<Request>
+    readonly body: { readonly from: (request: Request) => Effect.Effect<Body, AIError> }
+    readonly decode: (
+      response: HttpClientResponse.HttpClientResponse,
+      context: DecodeContext<Request>,
+    ) => Effect.Effect<Started<Token>, AIError>
+  }
+  readonly status: {
+    readonly path: (token: Token) => string
+    readonly decode: (
+      response: HttpClientResponse.HttpClientResponse,
+      context: PollContext<Token>,
+    ) => Effect.Effect<Snapshot, AIError>
+  }
+  readonly result: {
+    readonly path: (token: Token) => string
+    readonly decode: (
+      response: HttpClientResponse.HttpClientResponse,
+      context: PollContext<Token>,
+    ) => Effect.Effect<Response, AIError>
+  }
+  readonly cancel?: {
+    readonly method: AuthInput["method"]
+    readonly path: (token: Token) => string
+  }
+}
+
+export const queued = <Request, Response, Token>(
+  input: Omit<Queued<Request, Response, Token>, "kind">,
+): Queued<Request, Response, Token> => ({
+  kind: "queued",
+  ...input,
+})
+
+export type Mode = "generate" | "stream"
+
+export type Addressed<Request> = Request & { readonly mode: Mode }
+
+export interface ResponseContext<Request> extends DecodeContext<Addressed<Request>> {
+  readonly http: HttpContext
+}
+
+/**
+ * One request whose body is parsed incrementally, like LLM protocols: `frames` → `step`* → `finish`. `generate` and
+ * `stream` share this state machine; `request.mode` lets a protocol pick a different body, path, or framing.
+ */
+export interface Streamed<Request, Event, Frame, State> {
+  readonly kind: "stream"
+  readonly id: string
+  readonly name: string
+  /** Common request fields this protocol cannot lower; the route rejects them before `body.from` runs. */
+  readonly unsupported?: ReadonlyArray<keyof Request & string>
+  readonly body: { readonly from: (request: Addressed<Request>) => Effect.Effect<Body, AIError> }
+  readonly frames: (
+    bytes: Stream.Stream<Uint8Array, AIError>,
+    context: DecodeContext<Addressed<Request>>,
+  ) => Stream.Stream<Frame, AIError>
+  readonly initial: () => State
+  readonly step: (state: State, frame: Frame) => Effect.Effect<readonly [State, ReadonlyArray<Event>], AIError>
+  /** Emit exactly one terminal event, or fail when the provider stopped before completing. */
+  readonly finish: (state: State, context: ResponseContext<Request>) => Effect.Effect<ReadonlyArray<Event>, AIError>
+}
+
+export const stream = <Request, Event, Frame, State>(
+  input: Omit<Streamed<Request, Event, Frame, State>, "kind">,
+): Streamed<Request, Event, Frame, State> => ({
+  kind: "stream",
+  ...input,
+})
+
 // ---------------------------------------------------------------------------
 // Response helpers
 // ---------------------------------------------------------------------------
@@ -52,7 +188,11 @@ export const inline = <Request, Response>(
 const context = (response: HttpClientResponse.HttpClientResponse) =>
  new HttpContext({ url: response.request.url, status: response.status, headers: response.headers })

-/** Read a text body while retaining the original payload and HTTP context on every downstream error. */
+/**
+ * Read a text body while retaining the original payload and HTTP context on every downstream error. `invalid` is a
+ * malformed provider document; `ended` is a generation that reached a terminal status without output (`failed` is
+ * provider-side, `cancelled`/`expired` mean the result will never exist); `contentPolicy` is a moderated result.
+ */
 export const text = Effect.fn("MediaProtocol.text")(function* (
  route: string,
  name: string,
@@ -77,9 +217,19 @@ export const text = Effect.fn("MediaProtocol.text")(function* (
    http,
    invalid: (message: string, cause?: unknown) =>
      new AIError({ reason: new InvalidProviderOutputError({ route, message, body, http, cause }) }),
+    ended: (status: Exclude<Status, "queued" | "running" | "completed">, message: string) =>
+      new AIError({
+        reason:
+          status === "failed"
+            ? new ProviderInternalError({ message, body, http })
+            : new InvalidRequestError({ message, body, http }),
+      }),
+    contentPolicy: (message: string) => new AIError({ reason: new ContentPolicyError({ message, body, http }) }),
  }
 })

+export type Output = Effect.Success<ReturnType<typeof text>>
+
 /** Read and Schema-decode a JSON body. Decode failures keep the raw body as `reason.body`. */
 export const decodeJson = <A>(route: string, name: string, schema: Schema.Codec<A, unknown>) => {
  const decode = Schema.decodeUnknownEffect(Schema.fromJsonString(schema))
@@ -92,4 +242,54 @@ export const decodeJson = <A>(route: string, name: string, schema: Schema.Codec<
  })
 }

+/** Decode a submission response into the token and first snapshot. */
+export const decodeStarted = <A, Token>(
+  route: string,
+  name: string,
+  schema: Schema.Codec<A, unknown>,
+  started: (value: A) => Started<Token>,
+) => {
+  const decode = decodeJson(route, name, schema)
+  return (response: HttpClientResponse.HttpClientResponse) =>
+    decode(response).pipe(Effect.map((output) => started(output.value)))
+}
+
+/** Map a provider status string through the protocol's table; unknown values are an invalid provider document. */
+export const status = <Table extends Record<string, Status>>(
+  table: Table,
+  raw: string,
+  output: Output,
+): Effect.Effect<Status, AIError> => {
+  const normalized: Status | undefined = table[raw]
+  if (normalized === undefined) return Effect.fail(output.invalid(`Unknown generation status "${raw}"`))
+  return Effect.succeed(normalized)
+}
+
+export const frameError = (route: string, message: string, body?: string, cause?: unknown) =>
+  new AIError({ reason: new InvalidProviderOutputError({ route, message, body, cause }) })
+
+export const incomplete = (route: string) =>
+  new AIError({
+    reason: new InvalidProviderOutputError({
+      route,
+      message: "The provider response ended unexpectedly.",
+      classification: "incomplete-stream",
+    }),
+  })
+
+/** Schema-decode one JSON stream frame. Decode failures keep the frame as `reason.body`. */
+export const decodeFrame = <A>(route: string, name: string, schema: Schema.Codec<A, unknown>) => {
+  const decode = Schema.decodeUnknownEffect(Schema.fromJsonString(schema))
+  return (frame: string) =>
+    decode(frame).pipe(
+      Effect.mapError((cause) => frameError(route, `${name} sent an invalid stream event`, frame, cause)),
+    )
+}
+
+/** A `url` asset whose provider-declared retention window starts now. */
+export const expiringUrl = (url: string, retention: Duration.Duration, options?: Parameters<typeof Media.url>[1]) =>
+  Clock.currentTimeMillis.pipe(
+    Effect.map((now) => Media.url(url, { ...options, expiresAt: now + Duration.toMillis(retention) })),
+  )
+
 export * as MediaProtocol from "./media-protocol.js"
@@ -1,11 +1,20 @@
-import { Effect } from "effect"
-import { Headers, HttpClientRequest } from "effect/unstable/http"
-import { Auth } from "./auth.js"
+import { Effect, Schema, Stream } from "effect"
+import { Headers, HttpClientRequest, type HttpClientResponse } from "effect/unstable/http"
+import { Auth, type AuthInput } from "./auth.js"
 import { Endpoint } from "./endpoint.js"
 import type { Interface } from "./executor-service.js"
+import { RequestExecutor } from "./executor.js"
 import { MediaProtocol } from "./media-protocol.js"
+import { Generation, type Route as GenerationRoute } from "../generation.js"
 import { ProviderShared } from "../protocols/shared.js"
-import { AIError, HttpOptions, ProviderID, mergeHttpOptions } from "../schema/index.js"
+import {
+  AIError,
+  AIErrorReason,
+  HttpOptions,
+  InvalidRequestError,
+  ProviderID,
+  mergeHttpOptions,
+} from "../schema/index.js"
 import { sanitizeSurrogates } from "../utils/sanitize.js"

 export type Execute = Interface["execute"]
@@ -26,68 +35,328 @@ export interface ModelInput {
 }

 // ---------------------------------------------------------------------------
-// Route
+// Routes
 // ---------------------------------------------------------------------------

+/** One request, one response. */
 export interface Route<Request extends MediaRequest, Response> {
+  readonly kind: "inline"
  readonly id: string
  readonly provider: ProviderID
  readonly protocol: string
  readonly generate: (request: Request, execute: Execute) => Effect.Effect<Response, AIError>
 }

-export interface MakeInput<Request extends MediaRequest, Response> {
+/** Submit, then poll through the returned `Generation`. */
+export interface QueuedRoute<Request extends MediaRequest, Response> {
+  readonly kind: "queued"
+  readonly id: string
+  readonly provider: ProviderID
+  readonly protocol: string
+  readonly start: (request: Request, execute: Execute) => Effect.Effect<Generation<Response>, AIError>
+  /** Rebuild a handle from a persisted `Generation.token`; fails typed when the token is not this route's. */
+  readonly resume: (
+    model: MediaRequest["model"],
+    token: unknown,
+    execute: Execute,
+  ) => Effect.Effect<Generation<Response>, AIError>
+}
+
+/** One request whose response parses into events; `generate` runs the same stream and collects it. */
+export interface StreamRoute<Request extends MediaRequest, Event, Response> {
+  readonly kind: "stream"
+  readonly id: string
+  readonly provider: ProviderID
+  readonly protocol: string
+  readonly stream: (request: Request, execute: Execute) => Stream.Stream<Event, AIError>
+  readonly generate: (request: Request, execute: Execute) => Effect.Effect<Response, AIError>
+}
+
+export interface Composition<Request extends MediaRequest> {
  readonly id: string
  readonly provider: string | ProviderID
-  readonly protocol: MediaProtocol.Inline<Request, Response>
  readonly endpoint: Endpoint.Definition<MediaProtocol.Body, Request>
  readonly auth: Auth.Definition
  /** Deployment headers applied before transport authentication. */
  readonly headers?: Record<string, string>
 }

+export interface InlineInput<Request extends MediaRequest, Response> extends Composition<Request> {
+  readonly protocol: MediaProtocol.Inline<Request, Response>
+}
+
+export interface QueuedInput<Request extends MediaRequest, Response, Token> extends Composition<Request> {
+  readonly protocol: MediaProtocol.Queued<Request, Response, Token>
+}
+
+export interface StreamInput<Request extends MediaRequest, Event, Response, Frame, State>
+  extends Composition<MediaProtocol.Addressed<Request>> {
+  readonly protocol: MediaProtocol.Streamed<Request, Event, Frame, State>
+  readonly collect: (events: ReadonlyArray<Event>) => Effect.Effect<Response, AIError>
+}
+
 /**
 * Compose an inline media protocol with an endpoint and auth into a runnable route. The route owns the transport
- * plumbing every media protocol would otherwise duplicate: option merging, surrogate sanitizing, unsupported-field rejection, URL and query
- * rendering, auth headers, JSON vs multipart encoding, and handing the response back to the protocol for decoding.
+ * plumbing every media protocol would otherwise duplicate: option merging, surrogate sanitizing, unsupported-field
+ * rejection, URL and query rendering, auth headers, JSON, multipart, or binary encoding, and handing responses back
+ * to the protocol.
 */
-export const make = <Request extends MediaRequest, Response>(
-  input: MakeInput<Request, Response>,
+export const inline = <Request extends MediaRequest, Response>(
+  input: InlineInput<Request, Response>,
 ): Route<Request, Response> => {
+  const transport = makeTransport(input)
+  return {
+    kind: "inline",
+    id: input.id,
+    provider: transport.provider,
+    protocol: input.protocol.id,
+    generate: Effect.fn(`MediaRoute.generate`)(function* (request: Request, execute: Execute) {
+      const submitted = yield* transport.submit(
+        request,
+        { unsupported: input.protocol.unsupported, from: input.protocol.body.from },
+        execute,
+      )
+      return yield* input.protocol.response.decode(submitted.response, submitted.context)
+    }),
+  }
+}
+
+/**
+ * Compose a queued media protocol the same way, adding `start`/`resume` handles whose polls reuse the route's auth,
+ * deployment headers, and (for `start`) the request's `http` overlay. The token is decoded once at the boundary and
+ * closed over by the resulting `Generation.Route`.
+ */
+export const queued = <Request extends MediaRequest, Response, Token>(
+  input: QueuedInput<Request, Response, Token>,
+): QueuedRoute<Request, Response> => {
+  const transport = makeTransport(input)
+  const protocol = input.protocol
+  const decodeToken = Schema.decodeUnknownEffect(protocol.token)
+  // A protocol producing a token its own codec rejects is a programmer defect, not a provider error.
+  const encodeToken = Schema.encodeSync(protocol.token)
+
+  const generationRoute = (token: Token, http: HttpOptions | undefined, execute: Execute) => {
+    const poll = <A>(operation: {
+      readonly path: (token: Token) => string
+      readonly decode: (
+        response: HttpClientResponse.HttpClientResponse,
+        context: MediaProtocol.PollContext<Token>,
+      ) => Effect.Effect<A, AIError>
+    }) =>
+      transport
+        .call("GET", operation.path(token), http, execute)
+        .pipe(Effect.flatMap((sent) => operation.decode(sent.response, { token, auth: sent.auth })))
+    const cancel = protocol.cancel
+    const route: GenerationRoute<Response> = {
+      status: poll(protocol.status),
+      result: poll(protocol.result),
+      cancel:
+        cancel === undefined
+          ? undefined
+          : transport.call(cancel.method, cancel.path(token), http, execute).pipe(Effect.asVoid),
+    }
+    return route
+  }
+
+  const start = Effect.fn("MediaRoute.start")(function* (request: Request, execute: Execute) {
+    const submitted = yield* transport.submit(
+      request,
+      { unsupported: protocol.unsupported, prepare: protocol.start.prepare, from: protocol.start.body.from },
+      execute,
+    )
+    const started = yield* protocol.start.decode(submitted.response, submitted.context)
+    const route = generationRoute(started.token, submitted.context.request.http, execute)
+    return new Generation(route, encodeToken(started.token), started.snapshot)
+  })
+
+  const resume = Effect.fn("MediaRoute.resume")(function* (
+    model: MediaRequest["model"],
+    raw: unknown,
+    execute: Execute,
+  ) {
+    const token = yield* decodeToken(raw).pipe(
+      Effect.mapError(
+        (cause) =>
+          new AIError({
+            reason: new InvalidRequestError({
+              message: `${input.id} cannot resume a generation from this token`,
+              cause,
+            }),
+          }),
+      ),
+    )
+    const route = generationRoute(token, transport.http(model), execute)
+    return new Generation(route, encodeToken(token), yield* route.status)
+  })
+
+  return { kind: "queued", id: input.id, provider: transport.provider, protocol: protocol.id, start, resume }
+}
+
+/** Compose a streaming media protocol; `generate` runs the same stream in `generate` mode and folds it with `collect`. */
+export const stream = <Request extends MediaRequest, Event, Response, Frame, State>(
+  input: StreamInput<Request, Event, Response, Frame, State>,
+): StreamRoute<Request, Event, Response> => {
+  const transport = makeTransport(input)
+  const protocol = input.protocol
+  const events = (request: Request, execute: Execute, mode: MediaProtocol.Mode) =>
+    Stream.unwrap(
+      Effect.gen(function* () {
+        const submitted = yield* transport.submit(
+          { ...request, mode },
+          { unsupported: protocol.unsupported, from: protocol.body.from },
+          execute,
+        )
+        const http = RequestExecutor.responseHttp(submitted.response)
+        return Stream.suspend(() => {
+          // Parser state is local to one response, exactly like `Route.make`'s LLM stream loop.
+          let state = protocol.initial()
+          return protocol.frames(RequestExecutor.responseStream(submitted.response), submitted.context).pipe(
+            Stream.mapEffect((frame) =>
+              protocol.step(state, frame).pipe(
+                Effect.map(([next, output]) => {
+                  state = next
+                  return output
+                }),
+              ),
+            ),
+            Stream.flattenIterable,
+            Stream.concat(
+              Stream.suspend(() => Stream.fromIterableEffect(protocol.finish(state, { ...submitted.context, http }))),
+            ),
+            Stream.mapError((error) =>
+              error.reason.http !== undefined
+                ? error
+                : new AIError({
+                    reason: AIErrorReason.make({
+                      ...error.reason,
+                      message: error.reason.message,
+                      cause: error.reason.cause,
+                      http,
+                    }),
+                  }),
+            ),
+          )
+        })
+      }),
+    )
+  return {
+    kind: "stream",
+    id: input.id,
+    provider: transport.provider,
+    protocol: protocol.id,
+    stream: (request, execute) => events(request, execute, "stream"),
+    generate: (request, execute) =>
+      events(request, execute, "generate").pipe(Stream.runCollect, Effect.flatMap(input.collect)),
+  }
+}
+
+// ---------------------------------------------------------------------------
+// Transport plumbing shared by every kind
+// ---------------------------------------------------------------------------
+
+const makeTransport = <Request extends MediaRequest>(input: Composition<Request>) => {
  const provider = ProviderID.make(input.provider)
  const routeHttp = input.headers === undefined ? undefined : new HttpOptions({ headers: input.headers })
  const authorize = Auth.toEffect(input.auth)
+  const baseURL = (path: string) => new URL(`${ProviderShared.trimBaseUrl(input.endpoint.baseURL ?? "")}${path}`)
+  /** `auth` is only what `Auth` added, never deployment headers. */
+  const send = Effect.fn("MediaRoute.send")(function* (
+    call: {
+      readonly method: AuthInput["method"]
+      readonly url: URL
+      readonly headers: Headers.Headers
+      readonly request: AuthInput["request"]
+      readonly body?: MediaProtocol.Body
+    },
+    execute: Execute,
+  ) {
+    const encoded = encode(call.body, call.headers)
+    const url = call.url.toString()
+    const headers = yield* authorize({
+      request: call.request,
+      method: call.method,
+      url,
+      body: encoded.text,
+      headers: encoded.headers,
+    })
+    const response = yield* execute(
+      encoded.apply(HttpClientRequest.make(call.method)(url).pipe(HttpClientRequest.setHeaders(headers))),
+    )
+    return { response, auth: Object.fromEntries(Object.entries(headers).filter(([key]) => !(key in call.headers))) }
+  })
  return {
-    id: input.id,
    provider,
-    protocol: input.protocol.id,
-    generate: Effect.fn(`MediaRoute.generate`)(function* (request: Request, execute: Execute) {
-      yield* rejectUnsupported(input.id, provider, request, input.protocol.unsupported)
+    /** Route and model overlays; `start` additionally merges the request's own `http`. */
+    http: (model: MediaRequest["model"]) => mergeHttpOptions(routeHttp, model.http),
+    /** POST the protocol body to the route endpoint. */
+    submit: Effect.fn("MediaRoute.submit")(function* (
+      request: Request,
+      protocol: {
+        readonly unsupported?: ReadonlyArray<keyof Request & string>
+        readonly prepare?: MediaProtocol.Prepare<Request>
+        readonly from: (request: Request) => Effect.Effect<MediaProtocol.Body, AIError>
+      },
+      execute: Execute,
+    ) {
+      yield* rejectUnsupported(input.id, provider, request, protocol.unsupported)
      const http = mergeHttpOptions(routeHttp, request.model.http, request.http)
+      const headers = Headers.fromInput(http?.headers)
+      const prepared =
+        protocol.prepare === undefined
+          ? request
+          : yield* protocol.prepare(request, (path, body) =>
+              send({ method: "POST", url: baseURL(path), headers, request, body }, execute).pipe(
+                Effect.map((sent) => sent.response),
+              ),
+            )
      // Sanitize after merging so model-level overlays are covered; the model value is restored, not sanitized.
-      const resolved: Request = { ...sanitizeSurrogates({ ...request, http }), model: request.model }
-      const body = yield* input.protocol.body.from(resolved)
-      const url = Endpoint.render(input.endpoint, { request: resolved, body })
-      for (const [key, value] of Object.entries(http?.query ?? {})) url.searchParams.set(key, value)
-      const encoded = body.type === "json" ? ProviderShared.encodeJson(body.value) : "[multipart/form-data]"
-      const baseHeaders = Headers.fromInput(http?.headers)
-      const headers = yield* authorize({
-        request: resolved,
-        method: "POST",
-        url: url.toString(),
-        body: encoded,
-        // The HTTP client sets the multipart boundary; a caller-supplied content-type would corrupt it.
-        headers: body.type === "multipart" ? Headers.remove(baseHeaders, "content-type") : baseHeaders,
-      })
-      const transport = HttpClientRequest.post(url.toString()).pipe(
-        HttpClientRequest.setHeaders(headers),
-        body.type === "json"
-          ? HttpClientRequest.bodyText(encoded, "application/json")
-          : HttpClientRequest.bodyFormData(body.value),
+      const resolved: Request = { ...sanitizeSurrogates({ ...prepared, http }), model: request.model }
+      const body = yield* protocol.from(resolved)
+      const url = withQuery(
+        withQuery(
+          Endpoint.render(input.endpoint, { request: resolved, body }),
+          body.type === "multipart" ? undefined : body.query,
+        ),
+        http?.query,
      )
-      const response = yield* execute(transport)
-      return yield* input.protocol.response.decode(response, { request: resolved, body })
+      const sent = yield* send({ method: "POST", url, headers, request: resolved, body }, execute)
+      return { response: sent.response, context: { request: resolved, body } }
    }),
+    /** Bodiless follow-up call (status, result, cancel) with the same auth and headers as `submit`. */
+    call: (method: AuthInput["method"], path: string, http: HttpOptions | undefined, execute: Execute) => {
+      // Provider-issued absolute URLs (fal `status_url`) are used as-is; everything else resolves against the base.
+      const url = withQuery(/^https?:\/\//.test(path) ? new URL(path) : baseURL(path), http?.query)
+      for (const [key, value] of Object.entries(input.endpoint.query ?? {})) url.searchParams.set(key, value)
+      return send({ method, url, headers: Headers.fromInput(http?.headers), request: { http } }, execute)
+    },
+  }
+}
+
+const withQuery = (url: URL, query: MediaProtocol.Query | undefined) => {
+  for (const [key, value] of Object.entries(query ?? {})) {
+    url.searchParams.delete(key)
+    for (const item of typeof value === "string" ? [value] : value) url.searchParams.append(key, item)
+  }
+  return url
+}
+
+const encode = (body: MediaProtocol.Body | undefined, headers: Headers.Headers) => {
+  if (body === undefined) return { text: "", headers, apply: (request: HttpClientRequest.HttpClientRequest) => request }
+  if (body.type === "json") {
+    const text = ProviderShared.encodeJson(body.value)
+    return { text, headers, apply: HttpClientRequest.bodyText(text, "application/json") }
+  }
+  if (body.type === "binary")
+    return {
+      text: `[${body.contentType}]`,
+      headers,
+      apply: HttpClientRequest.bodyUint8Array(body.value, body.contentType),
+    }
+  return {
+    text: "[multipart/form-data]",
+    // The HTTP client sets the multipart boundary; a caller-supplied content-type would corrupt it.
+    headers: Headers.remove(headers, "content-type"),
+    apply: HttpClientRequest.bodyFormData(body.value),
  }
 }

@@ -0,0 +1,51 @@
+import { Context, Effect, Layer, Stream } from "effect"
+import { RequestExecutor } from "./route/executor.js"
+import type { AIError } from "./schema/index.js"
+import type { SpeechEvent, SpeechOptions, SpeechRequestFor, SpeechResponse } from "./speech.js"
+
+export interface Interface {
+  readonly generate: <Options extends SpeechOptions>(
+    request: SpeechRequestFor<Options>,
+  ) => Effect.Effect<SpeechResponse, AIError>
+  readonly stream: <Options extends SpeechOptions>(
+    request: SpeechRequestFor<Options>,
+  ) => Stream.Stream<SpeechEvent, AIError>
+}
+
+export class Service extends Context.Service<Service, Interface>()("@opencode/SpeechClient") {}
+
+export const generate = <Options extends SpeechOptions>(
+  request: SpeechRequestFor<Options>,
+): Effect.Effect<SpeechResponse, AIError, Service> =>
+  Effect.gen(function* () {
+    const client = yield* Service
+    return yield* client.generate(request)
+  })
+
+export const stream = <Options extends SpeechOptions>(
+  request: SpeechRequestFor<Options>,
+): Stream.Stream<SpeechEvent, AIError, Service> =>
+  Stream.unwrap(
+    Effect.gen(function* () {
+      const client = yield* Service
+      return client.stream(request)
+    }),
+  )
+
+export const layer: Layer.Layer<Service, never, RequestExecutor.Service> = Layer.effect(
+  Service,
+  Effect.gen(function* () {
+    const executor = yield* RequestExecutor.Service
+    return Service.of({
+      generate: (request) => request.model.route.generate(request, executor.execute),
+      stream: (request) => request.model.route.stream(request, executor.execute),
+    })
+  }),
+)
+
+export const SpeechClient = {
+  Service,
+  layer,
+  generate,
+  stream,
+} as const
@@ -0,0 +1,217 @@
+import { Effect, Schema, Stream } from "effect"
+import { Media } from "./media.js"
+import { MediaModel, composeRoute, tryRequest } from "./media-model.js"
+import { MediaRoute } from "./route/media.js"
+import type { MediaProtocol } from "./route/media-protocol.js"
+import { AIError, HttpOptions, MediaUsage, ProviderMetadata } from "./schema/index.js"
+import { SpeechClient, Service } from "./speech-client.js"
+
+// ---------------------------------------------------------------------------
+// Model
+// ---------------------------------------------------------------------------
+
+export type SpeechOptions = Record<string, unknown>
+
+export type SpeechRoute<Options extends SpeechOptions = SpeechOptions> = MediaRoute.StreamRoute<
+  SpeechRequestFor<Options>,
+  SpeechEvent,
+  SpeechResponse
+>
+
+export class SpeechModel<Options extends SpeechOptions = SpeechOptions> extends MediaModel<
+  SpeechRoute<Options>,
+  Options
+> {
+  declare protected readonly _SpeechModel: void
+
+  static make<Options extends SpeechOptions = SpeechOptions>(input: MediaModel.Input<SpeechRoute<Options>>) {
+    return new SpeechModel<Options>(input)
+  }
+
+  /** Compose a streaming speech protocol with its canonical path into a model for one deployment. */
+  static fromRoute<Options extends SpeechOptions = SpeechOptions, Frame = unknown, State = unknown>(
+    route: SpeechModel.RouteInput<Options, Frame, State>,
+    input: MediaRoute.ModelInput,
+  ) {
+    return new SpeechModel<Options>({
+      id: input.id,
+      provider: route.provider,
+      http: input.http,
+      route: composeRoute(
+        (composition) => MediaRoute.stream({ ...composition, collect: collectResponse }),
+        route,
+        input,
+      ),
+    })
+  }
+}
+
+export namespace SpeechModel {
+  export type RouteInput<
+    Options extends SpeechOptions = SpeechOptions,
+    Frame = unknown,
+    State = unknown,
+  > = MediaModel.RouteInput<
+    MediaProtocol.Addressed<SpeechRequestFor<Options>>,
+    MediaProtocol.Streamed<SpeechRequestFor<Options>, SpeechEvent, Frame, State>
+  >
+}
+
+export const SpeechModelSchema = Schema.declare((value): value is SpeechModel => value instanceof SpeechModel, {
+  expected: "Speech.Model",
+})
+
+// ---------------------------------------------------------------------------
+// Request
+// ---------------------------------------------------------------------------
+
+/** Provider-native: a name on OpenAI and Gemini, a voice id on ElevenLabs and Cartesia; `{ id }` is an OpenAI custom voice. */
+export const SpeechVoice = Schema.Union([Schema.String, Schema.Struct({ id: Schema.String })]).annotate({
+  identifier: "Speech.Voice",
+})
+export type SpeechVoice = Schema.Schema.Type<typeof SpeechVoice>
+
+export type SpeechFormat = "mp3" | "wav" | "pcm" | "opus" | "aac" | "flac" | (string & {})
+
+/** Granularity is provider-native: characters on ElevenLabs, words on Cartesia. */
+export const SpeechTimestamp = Schema.Struct({
+  text: Schema.String,
+  startSeconds: Schema.Number,
+  endSeconds: Schema.Number,
+}).annotate({ identifier: "Speech.Timestamp" })
+export type SpeechTimestamp = Schema.Schema.Type<typeof SpeechTimestamp>
+
+export class SpeechRequest extends Schema.Class<SpeechRequest>("Speech.Request")({
+  model: SpeechModelSchema,
+  text: Schema.String,
+  voice: Schema.optional(SpeechVoice),
+  format: Schema.optional(Schema.String),
+  speed: Schema.optional(Schema.Number),
+  language: Schema.optional(Schema.String),
+  instructions: Schema.optional(Schema.String),
+  timestamps: Schema.optional(Schema.Boolean),
+  providerOptions: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
+  http: Schema.optional(HttpOptions),
+}) {
+  declare protected readonly _SpeechRequest: void
+}
+
+export type SpeechRequestFor<Options extends SpeechOptions = SpeechOptions> = Omit<
+  SpeechRequest,
+  "model" | "providerOptions"
+> & {
+  readonly model: SpeechModel<Options>
+  readonly providerOptions?: Options
+}
+
+export type SpeechModelOptions<Model> = Model extends SpeechModel<infer Options> ? Options : never
+
+export type SpeechRequestInput<Model extends SpeechModel = SpeechModel> = Omit<
+  ConstructorParameters<typeof SpeechRequest>[0],
+  "model" | "providerOptions" | "http" | "format"
+> & {
+  readonly model: Model
+  readonly format?: SpeechFormat
+  readonly providerOptions?: NoInfer<SpeechModelOptions<Model>>
+  readonly http?: HttpOptions.Input
+}
+
+// ---------------------------------------------------------------------------
+// Response and events
+// ---------------------------------------------------------------------------
+
+export class SpeechResponse extends Schema.Class<SpeechResponse>("Speech.Response")({
+  /** The complete audio. Headerless PCM carries `info.encoding`, `info.sampleRate`, and `info.channels`. */
+  audio: Media.AssetSchema,
+  timestamps: Schema.optional(Schema.Array(SpeechTimestamp)),
+  usage: Schema.optional(MediaUsage),
+  notices: Schema.optional(Schema.Array(Media.Notice)),
+  providerMetadata: Schema.optional(ProviderMetadata),
+}) {}
+
+export const SpeechAudioDeltaEvent = Schema.Struct({
+  type: Schema.tag("audio-delta"),
+  chunk: Schema.Uint8Array,
+}).annotate({ identifier: "Speech.Event.AudioDelta" })
+
+export const SpeechTimestampsEvent = Schema.Struct({
+  type: Schema.tag("timestamps"),
+  items: Schema.Array(SpeechTimestamp),
+}).annotate({ identifier: "Speech.Event.Timestamps" })
+
+/** `audio` is every `audio-delta` chunk concatenated, so the route holds the whole clip in memory until `finish`. */
+export const SpeechFinishEvent = Schema.Struct({
+  type: Schema.tag("finish"),
+  audio: Media.AssetSchema,
+  usage: Schema.optional(MediaUsage),
+  notices: Schema.optional(Schema.Array(Media.Notice)),
+  providerMetadata: Schema.optional(ProviderMetadata),
+}).annotate({ identifier: "Speech.Event.Finish" })
+
+const speechEventTagged = Schema.Union([SpeechAudioDeltaEvent, SpeechTimestampsEvent, SpeechFinishEvent]).pipe(
+  Schema.toTaggedUnion("type"),
+)
+export const SpeechEvent = Object.assign(speechEventTagged, {
+  is: {
+    audioDelta: speechEventTagged.guards["audio-delta"],
+    timestamps: speechEventTagged.guards.timestamps,
+    finish: speechEventTagged.guards.finish,
+  },
+})
+export type SpeechEvent = Schema.Schema.Type<typeof speechEventTagged>
+
+const collectResponse = (events: ReadonlyArray<SpeechEvent>): Effect.Effect<SpeechResponse> => {
+  const finish = events.find(SpeechEvent.is.finish)
+  // Every speech protocol's `finish` emits the terminal event or fails, so a completed stream always has one.
+  if (finish === undefined) return Effect.die(new Error("The speech stream completed without a finish event"))
+  const timestamps = events.filter(SpeechEvent.is.timestamps).flatMap((event) => event.items)
+  return Effect.succeed(
+    new SpeechResponse({
+      audio: finish.audio,
+      timestamps: timestamps.length === 0 ? undefined : timestamps,
+      usage: finish.usage,
+      notices: finish.notices,
+      providerMetadata: finish.providerMetadata,
+    }),
+  )
+}
+
+// ---------------------------------------------------------------------------
+// Request-shaped call API
+// ---------------------------------------------------------------------------
+
+export function request<const Model extends SpeechModel>(
+  input: SpeechRequestInput<Model>,
+): SpeechRequestFor<SpeechModelOptions<Model>>
+export function request(input: SpeechRequest): SpeechRequest
+export function request(input: SpeechRequest | SpeechRequestInput) {
+  if (input instanceof SpeechRequest) return input
+  return new SpeechRequest({
+    ...input,
+    http: input.http === undefined ? undefined : HttpOptions.make(input.http),
+  })
+}
+
+const requestEffect = (input: SpeechRequest | SpeechRequestInput) => tryRequest(() => request(input))
+
+export function generate<const Model extends SpeechModel>(
+  input: SpeechRequestInput<Model>,
+): Effect.Effect<SpeechResponse, AIError, Service>
+export function generate(input: SpeechRequest): Effect.Effect<SpeechResponse, AIError, Service>
+export function generate(input: SpeechRequest | SpeechRequestInput) {
+  return requestEffect(input).pipe(Effect.flatMap((request) => SpeechClient.generate(request)))
+}
+
+export function stream<const Model extends SpeechModel>(
+  input: SpeechRequestInput<Model>,
+): Stream.Stream<SpeechEvent, AIError, Service>
+export function stream(input: SpeechRequest): Stream.Stream<SpeechEvent, AIError, Service>
+export function stream(input: SpeechRequest | SpeechRequestInput) {
+  return Stream.unwrap(requestEffect(input).pipe(Effect.map((request) => SpeechClient.stream(request))))
+}
+
+export const Speech = {
+  request,
+  generate,
+  stream,
+} as const
@@ -0,0 +1,122 @@
+import { Context, Effect, Layer, Stream } from "effect"
+import { resultEvents, type AwaitOptions, type Generation } from "./generation.js"
+import { ProviderShared } from "./protocols/shared.js"
+import { RequestExecutor } from "./route/executor.js"
+import type { AIError } from "./schema/index.js"
+import {
+  responseEvents,
+  type TranscriptionEvent,
+  type TranscriptionModel,
+  type TranscriptionOptions,
+  type TranscriptionRequestFor,
+  type TranscriptionResponse,
+  type TranscriptionRoute,
+} from "./transcription.js"
+
+export interface Interface {
+  readonly generate: <Options extends TranscriptionOptions>(
+    request: TranscriptionRequestFor<Options>,
+    options?: AwaitOptions,
+  ) => Effect.Effect<TranscriptionResponse, AIError>
+  readonly stream: <Options extends TranscriptionOptions>(
+    request: TranscriptionRequestFor<Options>,
+    options?: AwaitOptions,
+  ) => Stream.Stream<TranscriptionEvent, AIError>
+  readonly start: <Options extends TranscriptionOptions>(
+    request: TranscriptionRequestFor<Options>,
+  ) => Effect.Effect<Generation<TranscriptionResponse>, AIError>
+  readonly resume: <Options extends TranscriptionOptions>(
+    model: TranscriptionModel<Options>,
+    token: unknown,
+  ) => Effect.Effect<Generation<TranscriptionResponse>, AIError>
+}
+
+export class Service extends Context.Service<Service, Interface>()("@opencode/TranscriptionClient") {}
+
+export const generate = <Options extends TranscriptionOptions>(
+  request: TranscriptionRequestFor<Options>,
+  options?: AwaitOptions,
+): Effect.Effect<TranscriptionResponse, AIError, Service> =>
+  Effect.gen(function* () {
+    const client = yield* Service
+    return yield* client.generate(request, options)
+  })
+
+export const stream = <Options extends TranscriptionOptions>(
+  request: TranscriptionRequestFor<Options>,
+  options?: AwaitOptions,
+): Stream.Stream<TranscriptionEvent, AIError, Service> =>
+  Stream.unwrap(
+    Effect.gen(function* () {
+      const client = yield* Service
+      return client.stream(request, options)
+    }),
+  )
+
+export const start = <Options extends TranscriptionOptions>(
+  request: TranscriptionRequestFor<Options>,
+): Effect.Effect<Generation<TranscriptionResponse>, AIError, Service> =>
+  Effect.gen(function* () {
+    const client = yield* Service
+    return yield* client.start(request)
+  })
+
+export const resume = <Options extends TranscriptionOptions>(
+  model: TranscriptionModel<Options>,
+  token: unknown,
+): Effect.Effect<Generation<TranscriptionResponse>, AIError, Service> =>
+  Effect.gen(function* () {
+    const client = yield* Service
+    return yield* client.resume(model, token)
+  })
+
+const notQueued = <Options extends TranscriptionOptions>(route: TranscriptionRoute<Options>, operation: string) =>
+  ProviderShared.unsupportedOperation({
+    operation: `transcription.${operation}`,
+    provider: route.provider,
+    route: route.id,
+    message: `${route.provider}/${route.id} is not a queued route; use Transcription.generate or Transcription.stream`,
+  })
+
+export const layer: Layer.Layer<Service, never, RequestExecutor.Service> = Layer.effect(
+  Service,
+  Effect.gen(function* () {
+    const executor = yield* RequestExecutor.Service
+    const start = <Options extends TranscriptionOptions>(request: TranscriptionRequestFor<Options>) => {
+      const route = request.model.route
+      if (route.kind !== "queued") return Effect.fail(notQueued(route, "start"))
+      return route.start(request, executor.execute)
+    }
+    return Service.of({
+      start,
+      resume: (model, token) => {
+        const route = model.route
+        if (route.kind !== "queued") return Effect.fail(notQueued(route, "resume"))
+        return route.resume(model, token, executor.execute)
+      },
+      generate: (request, options) => {
+        const route = request.model.route
+        if (route.kind !== "queued") return route.generate(request, executor.execute)
+        return start(request).pipe(Effect.flatMap((generation) => generation.await(options)))
+      },
+      stream: (request, options) => {
+        const route = request.model.route
+        if (route.kind === "stream") return route.stream(request, executor.execute)
+        if (route.kind === "queued")
+          return Stream.unwrap(
+            start(request).pipe(Effect.map((generation) => resultEvents(generation, responseEvents, options))),
+          )
+        return Stream.fromIterableEffect(Effect.map(route.generate(request, executor.execute), responseEvents))
+      },
+    })
+  }),
+)
+
+export const TranscriptionClient = {
+  Service,
+  layer,
+  generate,
+  stream,
+  start,
+  resume,
+} as const
@@ -0,0 +1,294 @@
+import { Effect, Schema, Stream } from "effect"
+import { Generation, ProgressEvent, QueuedEvent, type AwaitOptions } from "./generation.js"
+import { Media } from "./media.js"
+import { MediaModel, composeRoute, tryRequest } from "./media-model.js"
+import { MediaRoute } from "./route/media.js"
+import type { MediaProtocol } from "./route/media-protocol.js"
+import { AIError, HttpOptions, MediaUsage, ProviderMetadata } from "./schema/index.js"
+import { TranscriptionClient, Service } from "./transcription-client.js"
+
+// ---------------------------------------------------------------------------
+// Model
+// ---------------------------------------------------------------------------
+
+export type TranscriptionOptions = Record<string, unknown>
+
+export type TranscriptionRoute<Options extends TranscriptionOptions = TranscriptionOptions> =
+  | MediaRoute.Route<TranscriptionRequestFor<Options>, TranscriptionResponse>
+  | MediaRoute.StreamRoute<TranscriptionRequestFor<Options>, TranscriptionEvent, TranscriptionResponse>
+  | MediaRoute.QueuedRoute<TranscriptionRequestFor<Options>, TranscriptionResponse>
+
+export class TranscriptionModel<Options extends TranscriptionOptions = TranscriptionOptions> extends MediaModel<
+  TranscriptionRoute<Options>,
+  Options
+> {
+  declare protected readonly _TranscriptionModel: void
+
+  static make<Options extends TranscriptionOptions = TranscriptionOptions>(
+    input: MediaModel.Input<TranscriptionRoute<Options>>,
+  ) {
+    return new TranscriptionModel<Options>(input)
+  }
+
+  /** The number of type arguments selects the kind: `<Options>`, `<Options, Frame, State>`, or `<Options, Token>`. */
+  static fromRoute<Options extends TranscriptionOptions>(
+    route: TranscriptionModel.InlineRouteInput<Options>,
+    input: MediaRoute.ModelInput,
+  ): TranscriptionModel<Options>
+  static fromRoute<Options extends TranscriptionOptions, Frame, State>(
+    route: TranscriptionModel.StreamRouteInput<Options, Frame, State>,
+    input: MediaRoute.ModelInput,
+  ): TranscriptionModel<Options>
+  static fromRoute<Options extends TranscriptionOptions, Token>(
+    route: TranscriptionModel.QueuedRouteInput<Options, Token>,
+    input: MediaRoute.ModelInput,
+  ): TranscriptionModel<Options>
+  static fromRoute<Options extends TranscriptionOptions, Frame, State, Token>(
+    route: TranscriptionModel.RouteInput<Options, Frame, State, Token>,
+    input: MediaRoute.ModelInput,
+  ) {
+    const composed: TranscriptionRoute<Options> = isStreamInput(route)
+      ? composeRoute((composition) => MediaRoute.stream({ ...composition, collect: collectResponse }), route, input)
+      : isQueuedInput(route)
+        ? composeRoute(MediaRoute.queued, route, input)
+        : composeRoute(MediaRoute.inline, route, input)
+    return new TranscriptionModel<Options>({
+      id: input.id,
+      provider: route.provider,
+      http: input.http,
+      route: composed,
+    })
+  }
+}
+
+export namespace TranscriptionModel {
+  export type InlineRouteInput<Options extends TranscriptionOptions = TranscriptionOptions> = MediaModel.RouteInput<
+    TranscriptionRequestFor<Options>,
+    MediaProtocol.Inline<TranscriptionRequestFor<Options>, TranscriptionResponse>
+  >
+
+  export type StreamRouteInput<
+    Options extends TranscriptionOptions = TranscriptionOptions,
+    Frame = unknown,
+    State = unknown,
+  > = MediaModel.RouteInput<
+    MediaProtocol.Addressed<TranscriptionRequestFor<Options>>,
+    MediaProtocol.Streamed<TranscriptionRequestFor<Options>, TranscriptionEvent, Frame, State>
+  >
+
+  export type QueuedRouteInput<
+    Options extends TranscriptionOptions = TranscriptionOptions,
+    Token = unknown,
+  > = MediaModel.RouteInput<
+    TranscriptionRequestFor<Options>,
+    MediaProtocol.Queued<TranscriptionRequestFor<Options>, TranscriptionResponse, Token>
+  >
+
+  export type RouteInput<
+    Options extends TranscriptionOptions = TranscriptionOptions,
+    Frame = unknown,
+    State = unknown,
+    Token = unknown,
+  > = InlineRouteInput<Options> | StreamRouteInput<Options, Frame, State> | QueuedRouteInput<Options, Token>
+}
+
+const isStreamInput = <Options extends TranscriptionOptions, Frame, State, Token>(
+  route: TranscriptionModel.RouteInput<Options, Frame, State, Token>,
+): route is TranscriptionModel.StreamRouteInput<Options, Frame, State> => route.protocol.kind === "stream"
+
+const isQueuedInput = <Options extends TranscriptionOptions, Frame, State, Token>(
+  route: TranscriptionModel.RouteInput<Options, Frame, State, Token>,
+): route is TranscriptionModel.QueuedRouteInput<Options, Token> => route.protocol.kind === "queued"
+
+export const TranscriptionModelSchema = Schema.declare(
+  (value): value is TranscriptionModel => value instanceof TranscriptionModel,
+  { expected: "Transcription.Model" },
+)
+
+// ---------------------------------------------------------------------------
+// Request
+// ---------------------------------------------------------------------------
+
+export const TranscriptionTimestamps = Schema.Literals(["none", "segment", "word"])
+export type TranscriptionTimestamps = Schema.Schema.Type<typeof TranscriptionTimestamps>
+
+export class TranscriptionRequest extends Schema.Class<TranscriptionRequest>("Transcription.Request")({
+  model: TranscriptionModelSchema,
+  audio: Media.AssetSchema,
+  language: Schema.optional(Schema.String),
+  prompt: Schema.optional(Schema.String),
+  /** Routes that cannot produce the requested granularity fail typed; routes may return more than asked. */
+  timestamps: Schema.optional(TranscriptionTimestamps),
+  diarize: Schema.optional(Schema.Boolean),
+  speakers: Schema.optional(Schema.Int),
+  providerOptions: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
+  http: Schema.optional(HttpOptions),
+}) {
+  declare protected readonly _TranscriptionRequest: void
+}
+
+export type TranscriptionRequestFor<Options extends TranscriptionOptions = TranscriptionOptions> = Omit<
+  TranscriptionRequest,
+  "model" | "providerOptions"
+> & {
+  readonly model: TranscriptionModel<Options>
+  readonly providerOptions?: Options
+}
+
+export type TranscriptionModelOptions<Model> = Model extends TranscriptionModel<infer Options> ? Options : never
+
+export type TranscriptionRequestInput<Model extends TranscriptionModel = TranscriptionModel> = Omit<
+  ConstructorParameters<typeof TranscriptionRequest>[0],
+  "model" | "providerOptions" | "http"
+> & {
+  readonly model: Model
+  readonly providerOptions?: NoInfer<TranscriptionModelOptions<Model>>
+  readonly http?: HttpOptions.Input
+}
+
+// ---------------------------------------------------------------------------
+// Response and events
+// ---------------------------------------------------------------------------
+
+/** Speaker labels are provider-native (`A`, `0`, `spk:0`, or a known speaker name). */
+export const TranscriptionSegment = Schema.Struct({
+  text: Schema.String,
+  startSeconds: Schema.Number,
+  endSeconds: Schema.Number,
+  speaker: Schema.optional(Schema.String),
+}).annotate({ identifier: "Transcription.Segment" })
+export type TranscriptionSegment = Schema.Schema.Type<typeof TranscriptionSegment>
+
+export const TranscriptionWord = Schema.Struct({
+  text: Schema.String,
+  startSeconds: Schema.Number,
+  endSeconds: Schema.Number,
+  speaker: Schema.optional(Schema.String),
+  confidence: Schema.optional(Schema.Number),
+}).annotate({ identifier: "Transcription.Word" })
+export type TranscriptionWord = Schema.Schema.Type<typeof TranscriptionWord>
+
+const transcriptFields = {
+  text: Schema.String,
+  segments: Schema.optional(Schema.Array(TranscriptionSegment)),
+  words: Schema.optional(Schema.Array(TranscriptionWord)),
+  /** Provider-native language as detected or echoed (`en`, `english`, `en_us`), lowercased but not normalized. */
+  language: Schema.optional(Schema.String),
+  durationSeconds: Schema.optional(Schema.Number),
+  usage: Schema.optional(MediaUsage),
+  notices: Schema.optional(Schema.Array(Media.Notice)),
+  providerMetadata: Schema.optional(ProviderMetadata),
+}
+
+export class TranscriptionResponse extends Schema.Class<TranscriptionResponse>("Transcription.Response")(
+  transcriptFields,
+) {}
+
+export const TranscriptionTextDeltaEvent = Schema.Struct({
+  type: Schema.tag("text-delta"),
+  delta: Schema.String,
+}).annotate({ identifier: "Transcription.Event.TextDelta" })
+
+export const TranscriptionSegmentEvent = Schema.Struct({
+  type: Schema.tag("segment"),
+  segment: TranscriptionSegment,
+}).annotate({ identifier: "Transcription.Event.Segment" })
+
+export const TranscriptionFinishEvent = Schema.Struct({
+  type: Schema.tag("finish"),
+  ...transcriptFields,
+}).annotate({ identifier: "Transcription.Event.Finish" })
+
+const transcriptionEventTagged = Schema.Union([
+  QueuedEvent,
+  ProgressEvent,
+  TranscriptionTextDeltaEvent,
+  TranscriptionSegmentEvent,
+  TranscriptionFinishEvent,
+]).pipe(Schema.toTaggedUnion("type"))
+export const TranscriptionEvent = Object.assign(transcriptionEventTagged, {
+  is: {
+    generationQueued: transcriptionEventTagged.guards["generation-queued"],
+    generationProgress: transcriptionEventTagged.guards["generation-progress"],
+    textDelta: transcriptionEventTagged.guards["text-delta"],
+    segment: transcriptionEventTagged.guards.segment,
+    finish: transcriptionEventTagged.guards.finish,
+  },
+})
+export type TranscriptionEvent = Schema.Schema.Type<typeof transcriptionEventTagged>
+
+export const responseEvents = (response: TranscriptionResponse): ReadonlyArray<TranscriptionEvent> => [
+  TranscriptionFinishEvent.make({ ...response }),
+]
+
+const collectResponse = (events: ReadonlyArray<TranscriptionEvent>): Effect.Effect<TranscriptionResponse> => {
+  const finish = events.find(TranscriptionEvent.is.finish)
+  // Every transcription protocol's `finish` emits the terminal event or fails, so a completed stream always has one.
+  if (finish === undefined) return Effect.die(new Error("The transcription stream completed without a finish event"))
+  const { type: _type, ...transcript } = finish
+  return Effect.succeed(new TranscriptionResponse(transcript))
+}
+
+// ---------------------------------------------------------------------------
+// Request-shaped call API
+// ---------------------------------------------------------------------------
+
+export function request<const Model extends TranscriptionModel>(
+  input: TranscriptionRequestInput<Model>,
+): TranscriptionRequestFor<TranscriptionModelOptions<Model>>
+export function request(input: TranscriptionRequest): TranscriptionRequest
+export function request(input: TranscriptionRequest | TranscriptionRequestInput) {
+  if (input instanceof TranscriptionRequest) return input
+  return new TranscriptionRequest({
+    ...input,
+    http: input.http === undefined ? undefined : HttpOptions.make(input.http),
+  })
+}
+
+const requestEffect = (input: TranscriptionRequest | TranscriptionRequestInput) => tryRequest(() => request(input))
+
+export function generate<const Model extends TranscriptionModel>(
+  input: TranscriptionRequestInput<Model>,
+  options?: AwaitOptions,
+): Effect.Effect<TranscriptionResponse, AIError, Service>
+export function generate(
+  input: TranscriptionRequest,
+  options?: AwaitOptions,
+): Effect.Effect<TranscriptionResponse, AIError, Service>
+export function generate(input: TranscriptionRequest | TranscriptionRequestInput, options?: AwaitOptions) {
+  return requestEffect(input).pipe(Effect.flatMap((request) => TranscriptionClient.generate(request, options)))
+}
+
+export function stream<const Model extends TranscriptionModel>(
+  input: TranscriptionRequestInput<Model>,
+  options?: AwaitOptions,
+): Stream.Stream<TranscriptionEvent, AIError, Service>
+export function stream(
+  input: TranscriptionRequest,
+  options?: AwaitOptions,
+): Stream.Stream<TranscriptionEvent, AIError, Service>
+export function stream(input: TranscriptionRequest | TranscriptionRequestInput, options?: AwaitOptions) {
+  return Stream.unwrap(requestEffect(input).pipe(Effect.map((request) => TranscriptionClient.stream(request, options))))
+}
+
+/** Inline and streaming routes fail with `UnsupportedOperation`. */
+export function start<const Model extends TranscriptionModel>(
+  input: TranscriptionRequestInput<Model>,
+): Effect.Effect<Generation<TranscriptionResponse>, AIError, Service>
+export function start(input: TranscriptionRequest): Effect.Effect<Generation<TranscriptionResponse>, AIError, Service>
+export function start(input: TranscriptionRequest | TranscriptionRequestInput) {
+  return requestEffect(input).pipe(Effect.flatMap((request) => TranscriptionClient.start(request)))
+}
+
+export const resume = <Options extends TranscriptionOptions>(
+  model: TranscriptionModel<Options>,
+  token: unknown,
+): Effect.Effect<Generation<TranscriptionResponse>, AIError, Service> => TranscriptionClient.resume(model, token)
+
+export const Transcription = {
+  request,
+  generate,
+  stream,
+  start,
+  resume,
+} as const
@@ -0,0 +1,9 @@
+export const concatBytes = (chunks: ReadonlyArray<Uint8Array>) => {
+  if (chunks.length === 1) return chunks[0]
+  const bytes = new Uint8Array(chunks.reduce((total, chunk) => total + chunk.length, 0))
+  chunks.reduce((offset, chunk) => {
+    bytes.set(chunk, offset)
+    return offset + chunk.length
+  }, 0)
+  return bytes
+}
@@ -19,9 +19,12 @@ export const detectMediaType = (bytes: Uint8Array): string | undefined => {
  if (startsWith(bytes, [0x25, 0x50, 0x44, 0x46])) return "application/pdf"
  if (bytes.length >= 12 && ascii(bytes, 4, 8) === "ftyp") return "video/mp4"
  if (startsWith(bytes, [0x1a, 0x45, 0xdf, 0xa3])) return "video/webm"
-  if (startsWith(bytes, [0x49, 0x44, 0x33]) || startsWith(bytes, [0xff, 0xfb]) || startsWith(bytes, [0xff, 0xf3]))
-    return "audio/mpeg"
+  if (startsWith(bytes, [0x49, 0x44, 0x33])) return "audio/mpeg"
+  // An 11-bit frame sync; layer bits `00` mark AAC ADTS, any other layer is MPEG audio.
+  if (bytes.length >= 2 && bytes[0] === 0xff && (bytes[1] & 0xe0) === 0xe0)
+    return (bytes[1] & 0x06) === 0 ? "audio/aac" : "audio/mpeg"
  if (startsWith(bytes, [0x4f, 0x67, 0x67, 0x53])) return "audio/ogg"
+  if (startsWith(bytes, [0x66, 0x4c, 0x61, 0x43])) return "audio/flac"
  return undefined
 }

@@ -35,8 +38,11 @@ const EXTENSIONS: Readonly<Record<string, string>> = {
  mp4: "video/mp4",
  webm: "video/webm",
  mp3: "audio/mpeg",
+  m4a: "audio/mp4",
  wav: "audio/wav",
  ogg: "audio/ogg",
+  flac: "audio/flac",
+  aac: "audio/aac",
  txt: "text/plain",
  md: "text/markdown",
  csv: "text/csv",
@@ -44,3 +50,18 @@ const EXTENSIONS: Readonly<Record<string, string>> = {

 export const extensionMediaType = (path: string): string | undefined =>
  EXTENSIONS[path.slice(path.lastIndexOf(".") + 1).toLowerCase()]
+
+const EXTENSION_ALIASES: Readonly<Record<string, string>> = {
+  "audio/mp3": "mp3",
+  "audio/m4a": "m4a",
+  "audio/x-m4a": "m4a",
+  "audio/webm": "webm",
+  "audio/wave": "wav",
+  "audio/x-wav": "wav",
+  "audio/x-flac": "flac",
+}
+
+export const mediaTypeExtension = (mediaType: string): string | undefined => {
+  const type = mediaType.split(";", 1)[0].trim().toLowerCase()
+  return EXTENSION_ALIASES[type] ?? Object.entries(EXTENSIONS).find(([, known]) => known === type)?.[0]
+}
@@ -0,0 +1,96 @@
+import { Context, Effect, Layer, Stream } from "effect"
+import { resultEvents, type AwaitOptions, type Generation } from "./generation.js"
+import { RequestExecutor } from "./route/executor.js"
+import type { AIError } from "./schema/index.js"
+import {
+  responseEvents,
+  type VideoEvent,
+  type VideoModel,
+  type VideoOptions,
+  type VideoRequestFor,
+  type VideoResponse,
+} from "./video.js"
+
+export interface Interface {
+  readonly start: <Options extends VideoOptions>(
+    request: VideoRequestFor<Options>,
+  ) => Effect.Effect<Generation<VideoResponse>, AIError>
+  readonly resume: <Options extends VideoOptions>(
+    model: VideoModel<Options>,
+    token: unknown,
+  ) => Effect.Effect<Generation<VideoResponse>, AIError>
+  readonly generate: <Options extends VideoOptions>(
+    request: VideoRequestFor<Options>,
+    options?: AwaitOptions,
+  ) => Effect.Effect<VideoResponse, AIError>
+  readonly stream: <Options extends VideoOptions>(
+    request: VideoRequestFor<Options>,
+    options?: AwaitOptions,
+  ) => Stream.Stream<VideoEvent, AIError>
+}
+
+export class Service extends Context.Service<Service, Interface>()("@opencode/VideoClient") {}
+
+export const start = <Options extends VideoOptions>(
+  request: VideoRequestFor<Options>,
+): Effect.Effect<Generation<VideoResponse>, AIError, Service> =>
+  Effect.gen(function* () {
+    const client = yield* Service
+    return yield* client.start(request)
+  })
+
+export const resume = <Options extends VideoOptions>(
+  model: VideoModel<Options>,
+  token: unknown,
+): Effect.Effect<Generation<VideoResponse>, AIError, Service> =>
+  Effect.gen(function* () {
+    const client = yield* Service
+    return yield* client.resume(model, token)
+  })
+
+export const generate = <Options extends VideoOptions>(
+  request: VideoRequestFor<Options>,
+  options?: AwaitOptions,
+): Effect.Effect<VideoResponse, AIError, Service> =>
+  Effect.gen(function* () {
+    const client = yield* Service
+    return yield* client.generate(request, options)
+  })
+
+export const stream = <Options extends VideoOptions>(
+  request: VideoRequestFor<Options>,
+  options?: AwaitOptions,
+): Stream.Stream<VideoEvent, AIError, Service> =>
+  Stream.unwrap(
+    Effect.gen(function* () {
+      const client = yield* Service
+      return client.stream(request, options)
+    }),
+  )
+
+export const layer: Layer.Layer<Service, never, RequestExecutor.Service> = Layer.effect(
+  Service,
+  Effect.gen(function* () {
+    const executor = yield* RequestExecutor.Service
+    const start = <Options extends VideoOptions>(request: VideoRequestFor<Options>) =>
+      request.model.route.start(request, executor.execute)
+    return Service.of({
+      start,
+      resume: (model, token) => model.route.resume(model, token, executor.execute),
+      generate: (request, options) => start(request).pipe(Effect.flatMap((generation) => generation.await(options))),
+      stream: (request, options) =>
+        Stream.unwrap(
+          start(request).pipe(Effect.map((generation) => resultEvents(generation, responseEvents, options))),
+        ),
+    })
+  }),
+)
+
+export const VideoClient = {
+  Service,
+  layer,
+  start,
+  resume,
+  generate,
+  stream,
+} as const
@@ -0,0 +1,218 @@
+import { Effect, Schema, Stream } from "effect"
+import { Generation, ProgressEvent, QueuedEvent, type AwaitOptions } from "./generation.js"
+import { Media } from "./media.js"
+import { MediaModel, composeRoute, tryRequest } from "./media-model.js"
+import { MediaRoute } from "./route/media.js"
+import type { MediaProtocol } from "./route/media-protocol.js"
+import { AIError, HttpOptions, MediaUsage, ProviderMetadata } from "./schema/index.js"
+import { VideoClient, Service } from "./video-client.js"
+
+// ---------------------------------------------------------------------------
+// Model
+// ---------------------------------------------------------------------------
+
+export type VideoOptions = Record<string, unknown>
+
+export type VideoRoute<Options extends VideoOptions = VideoOptions> = MediaRoute.QueuedRoute<
+  VideoRequestFor<Options>,
+  VideoResponse
+>
+
+export class VideoModel<Options extends VideoOptions = VideoOptions> extends MediaModel<VideoRoute<Options>, Options> {
+  declare protected readonly _VideoModel: void
+
+  static make<Options extends VideoOptions = VideoOptions>(input: MediaModel.Input<VideoRoute<Options>>) {
+    return new VideoModel<Options>(input)
+  }
+
+  /** Compose a queued video protocol with its canonical start path into a model for one deployment. */
+  static fromRoute<Options extends VideoOptions = VideoOptions, Token = unknown>(
+    route: VideoModel.RouteInput<Options, Token>,
+    input: MediaRoute.ModelInput,
+  ) {
+    return new VideoModel<Options>({
+      id: input.id,
+      provider: route.provider,
+      http: input.http,
+      route: composeRoute(MediaRoute.queued, route, input),
+    })
+  }
+}
+
+export namespace VideoModel {
+  export type RouteInput<Options extends VideoOptions = VideoOptions, Token = unknown> = MediaModel.RouteInput<
+    VideoRequestFor<Options>,
+    MediaProtocol.Queued<VideoRequestFor<Options>, VideoResponse, Token>
+  >
+}
+
+export const VideoModelSchema = Schema.declare((value): value is VideoModel => value instanceof VideoModel, {
+  expected: "Video.Model",
+})
+
+// ---------------------------------------------------------------------------
+// Request
+// ---------------------------------------------------------------------------
+
+export type VideoAspectRatio = Media.AspectRatio
+export const VideoAspectRatio = Media.AspectRatio
+
+export type VideoResolution = "480p" | "720p" | "1080p" | "4k" | (string & {})
+
+/** Pinned frames. Routes that accept only a first frame fail typed when `last` is present. */
+export const VideoFrames = Schema.Struct({
+  first: Schema.optional(Media.AssetSchema),
+  last: Schema.optional(Media.AssetSchema),
+}).annotate({ identifier: "Video.Frames" })
+export type VideoFrames = Schema.Schema.Type<typeof VideoFrames>
+
+export class VideoRequest extends Schema.Class<VideoRequest>("Video.Request")({
+  model: VideoModelSchema,
+  prompt: Schema.String,
+  frames: Schema.optional(VideoFrames),
+  /** Style or subject references that guide the output without pinning a frame. */
+  references: Schema.optional(Schema.Array(Media.AssetSchema)),
+  /** Source video for edit or extension routes. */
+  video: Schema.optional(Media.AssetSchema),
+  durationSeconds: Schema.optional(Schema.Number),
+  aspectRatio: Schema.optional(VideoAspectRatio),
+  resolution: Schema.optional(Schema.String),
+  /** Whether to generate an audio track; routes whose audio is always on fail typed on `false`. */
+  audio: Schema.optional(Schema.Boolean),
+  n: Schema.optional(Schema.Int),
+  seed: Schema.optional(Schema.Number),
+  negativePrompt: Schema.optional(Schema.String),
+  providerOptions: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
+  http: Schema.optional(HttpOptions),
+}) {
+  declare protected readonly _VideoRequest: void
+}
+
+export type VideoRequestFor<Options extends VideoOptions = VideoOptions> = Omit<
+  VideoRequest,
+  "model" | "providerOptions"
+> & {
+  readonly model: VideoModel<Options>
+  readonly providerOptions?: Options
+}
+
+export type VideoModelOptions<Model> = Model extends VideoModel<infer Options> ? Options : never
+
+export type VideoRequestInput<Model extends VideoModel = VideoModel> = Omit<
+  ConstructorParameters<typeof VideoRequest>[0],
+  "model" | "providerOptions" | "http" | "resolution"
+> & {
+  readonly model: Model
+  readonly resolution?: VideoResolution
+  readonly providerOptions?: NoInfer<VideoModelOptions<Model>>
+  readonly http?: HttpOptions.Input
+}
+
+// ---------------------------------------------------------------------------
+// Response and events
+// ---------------------------------------------------------------------------
+
+export class VideoResponse extends Schema.Class<VideoResponse>("Video.Response")({
+  videos: Schema.Array(Media.AssetSchema),
+  usage: Schema.optional(MediaUsage),
+  notices: Schema.optional(Schema.Array(Media.Notice)),
+  providerMetadata: Schema.optional(ProviderMetadata),
+}) {
+  get video() {
+    return this.videos[0]
+  }
+}
+
+export const VideoOutputEvent = Schema.Struct({
+  type: Schema.tag("video"),
+  index: Schema.Number,
+  video: Media.AssetSchema,
+}).annotate({ identifier: "Video.Event.Video" })
+
+export const VideoFinishEvent = Schema.Struct({
+  type: Schema.tag("finish"),
+  usage: Schema.optional(MediaUsage),
+  notices: Schema.optional(Schema.Array(Media.Notice)),
+  providerMetadata: Schema.optional(ProviderMetadata),
+}).annotate({ identifier: "Video.Event.Finish" })
+
+const videoEventTagged = Schema.Union([QueuedEvent, ProgressEvent, VideoOutputEvent, VideoFinishEvent]).pipe(
+  Schema.toTaggedUnion("type"),
+)
+export const VideoEvent = Object.assign(videoEventTagged, {
+  is: {
+    generationQueued: videoEventTagged.guards["generation-queued"],
+    generationProgress: videoEventTagged.guards["generation-progress"],
+    video: videoEventTagged.guards.video,
+    finish: videoEventTagged.guards.finish,
+  },
+})
+export type VideoEvent = Schema.Schema.Type<typeof videoEventTagged>
+
+/** A completed response expanded into the streaming event shape. */
+export const responseEvents = (response: VideoResponse): ReadonlyArray<VideoEvent> => [
+  ...response.videos.map((video, index) => VideoOutputEvent.make({ index, video })),
+  VideoFinishEvent.make({
+    usage: response.usage,
+    notices: response.notices,
+    providerMetadata: response.providerMetadata,
+  }),
+]
+
+// ---------------------------------------------------------------------------
+// Request-shaped call API
+// ---------------------------------------------------------------------------
+
+export function request<const Model extends VideoModel>(
+  input: VideoRequestInput<Model>,
+): VideoRequestFor<VideoModelOptions<Model>>
+export function request(input: VideoRequest): VideoRequest
+export function request(input: VideoRequest | VideoRequestInput) {
+  if (input instanceof VideoRequest) return input
+  return new VideoRequest({
+    ...input,
+    http: input.http === undefined ? undefined : HttpOptions.make(input.http),
+  })
+}
+
+const requestEffect = (input: VideoRequest | VideoRequestInput) => tryRequest(() => request(input))
+
+export function start<const Model extends VideoModel>(
+  input: VideoRequestInput<Model>,
+): Effect.Effect<Generation<VideoResponse>, AIError, Service>
+export function start(input: VideoRequest): Effect.Effect<Generation<VideoResponse>, AIError, Service>
+export function start(input: VideoRequest | VideoRequestInput) {
+  return requestEffect(input).pipe(Effect.flatMap((request) => VideoClient.start(request)))
+}
+
+export function generate<const Model extends VideoModel>(
+  input: VideoRequestInput<Model>,
+  options?: AwaitOptions,
+): Effect.Effect<VideoResponse, AIError, Service>
+export function generate(input: VideoRequest, options?: AwaitOptions): Effect.Effect<VideoResponse, AIError, Service>
+export function generate(input: VideoRequest | VideoRequestInput, options?: AwaitOptions) {
+  return requestEffect(input).pipe(Effect.flatMap((request) => VideoClient.generate(request, options)))
+}
+
+/** Rebuild a generation handle from a persisted `Generation.token`, refreshing its status once. */
+export const resume = <Options extends VideoOptions>(
+  model: VideoModel<Options>,
+  token: unknown,
+): Effect.Effect<Generation<VideoResponse>, AIError, Service> => VideoClient.resume(model, token)
+
+export function stream<const Model extends VideoModel>(
+  input: VideoRequestInput<Model>,
+  options?: AwaitOptions,
+): Stream.Stream<VideoEvent, AIError, Service>
+export function stream(input: VideoRequest, options?: AwaitOptions): Stream.Stream<VideoEvent, AIError, Service>
+export function stream(input: VideoRequest | VideoRequestInput, options?: AwaitOptions) {
+  return Stream.unwrap(requestEffect(input).pipe(Effect.map((request) => VideoClient.stream(request, options))))
+}
+
+export const Video = {
+  request,
+  start,
+  generate,
+  resume,
+  stream,
+} as const
@@ -1,17 +1,38 @@
 import { describe, expect, test } from "bun:test"
-import { AIError, LanguageModel, LLM, LLMClient, Media, Provider } from "@opencode/ai"
+import {
+  AIError,
+  Generation,
+  Image,
+  LanguageModel,
+  LLM,
+  LLMClient,
+  Media,
+  Provider,
+  Speech,
+  SpeechClient,
+  SpeechEvent,
+  Video,
+  VideoClient,
+} from "@opencode/ai"
 import { Route, Protocol, WebSocketTransport } from "@opencode/ai/route"
 import { Provider as ProviderSubpath } from "@opencode/ai/provider"
 import {
+  AssemblyAI,
  Baseten,
+  Cartesia,
  CloudflareAIGateway,
  CloudflareWorkersAI,
+  Deepgram,
  DeepSeek,
+  ElevenLabs,
+  Fal,
  Fireworks,
+  Google,
  OpenCodeZen,
  OpenAI,
  OpenAICompatible,
  OpenRouter,
+  Runway,
  TypeSafeAI,
  VercelAIGateway,
  XAI,
@@ -36,6 +57,15 @@ describe("public exports", () => {
    expect(AIError).toBeFunction()
    expect(LanguageModel.make).toBeFunction()
    expect(Media.bytes).toBeFunction()
+    expect(Image.generate).toBeFunction()
+    expect(Video.start).toBeFunction()
+    expect(Video.resume).toBeFunction()
+    expect(VideoClient.layer).toBeDefined()
+    expect(Speech.generate).toBeFunction()
+    expect(Speech.stream).toBeFunction()
+    expect(SpeechClient.layer).toBeDefined()
+    expect(SpeechEvent.is.audioDelta).toBeFunction()
+    expect(Generation).toBeFunction()
    expect(Provider.make).toBeFunction()
    expect(ProviderSubpath.make).toBe(Provider.make)
    expect(TestLLM.layer).toBeFunction()
@@ -82,6 +112,21 @@ describe("public exports", () => {
    expect(XAI.provider.chat).toBe(XAI.chat)
    expect(XAI.configure({ apiKey: "fixture" }).responses("grok-4.3").route.id).toBe("openai-responses")
    expect(XAI.configure({ apiKey: "fixture" }).chat("grok-4.3").route.id).toBe("openai-compatible-chat")
+    expect(XAI.configure({ apiKey: "fixture" }).video("grok-imagine-video-1.5").route.id).toBe("xai-video")
+    expect(Fal.configure({ apiKey: "fixture" }).video("fal-ai/veo3.1").route.id).toBe("fal-video")
+    expect(Runway.configure({ apiKey: "fixture" }).video("gen4.5").route.id).toBe("runway-video")
+    expect(Runway.provider.video).toBe(Runway.video)
+    expect(OpenAI.configure({ apiKey: "fixture" }).speech("gpt-4o-mini-tts").route.id).toBe("openai-speech")
+    expect(Google.configure({ apiKey: "fixture" }).speech("gemini-2.5-flash-preview-tts").route.id).toBe(
+      "google-speech",
+    )
+    expect(ElevenLabs.configure({ apiKey: "fixture" }).speech("eleven_flash_v2_5").route.id).toBe("elevenlabs-speech")
+    expect(Cartesia.configure({ apiKey: "fixture" }).speech("sonic-3").route.id).toBe("cartesia-speech")
+    expect(Deepgram.configure({ apiKey: "fixture" }).speech("aura-2-thalia-en").route.id).toBe("deepgram-speech")
+    expect(OpenAI.configure({ apiKey: "fixture" }).transcription("gpt-transcribe").route.kind).toBe("stream")
+    expect(Google.configure({ apiKey: "fixture" }).transcription("gemini-3.5-transcribe").route.kind).toBe("stream")
+    expect(Deepgram.configure({ apiKey: "fixture" }).transcription("nova-3").route.kind).toBe("inline")
+    expect(AssemblyAI.configure({ apiKey: "fixture" }).transcription("universal-3-5-pro").route.kind).toBe("queued")
  })

  test("protocol barrels expose supported low-level routes", () => {
@@ -0,0 +1,98 @@
+{
+  "version": 1,
+  "metadata": {
+    "tags": [
+      "prefix:assemblyai-transcription",
+      "provider:assemblyai",
+      "protocol:assemblyai-transcription"
+    ],
+    "name": "assemblyai-transcription/uploads-submits-and-polls-a-transcript",
+    "recordedAt": "2026-09-23T14:03:07.504Z"
+  },
+  "interactions": [
+    {
+      "transport": "http",
+      "request": {
+        "method": "POST",
+        "url": "https://api.eu.assemblyai.com/v2/upload",
+        "headers": {
+          "content-type": "application/octet-stream"
+        },
+        "body": "[audio]"
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "application/json"
+        },
+        "body": "{\"upload_url\":\"https://cdn.assemblyai.com/upload/f057ab4f09d5f28527be4b830b3d92f8bf6a97e5f854837cf22efbf2d09b5182/ed6a04e2-e34c-4a70-89da-b1ec34d2fb5f\"}\n"
+      }
+    },
+    {
+      "transport": "http",
+      "request": {
+        "method": "POST",
+        "url": "https://api.eu.assemblyai.com/v2/transcript",
+        "headers": {
+          "content-type": "application/json"
+        },
+        "body": "{\"audio_url\":\"https://cdn.assemblyai.com/upload/f057ab4f09d5f28527be4b830b3d92f8bf6a97e5f854837cf22efbf2d09b5182/ed6a04e2-e34c-4a70-89da-b1ec34d2fb5f\",\"speech_models\":[\"universal-3-5-pro\"],\"language_detection\":true,\"speaker_labels\":true}"
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "application/json; charset=UTF-8"
+        },
+        "body": "{\"id\": \"a1de21ad-a178-4c28-b8f0-4b0334b72126\", \"language_model\": \"assemblyai_default\", \"acoustic_model\": \"assemblyai_default\", \"language_code\": null, \"speech_understanding\": null, \"translated_texts\": null, \"status\": \"processing\", \"audio_url\": \"https://cdn.assemblyai.com/upload/f057ab4f09d5f28527be4b830b3d92f8bf6a97e5f854837cf22efbf2d09b5182/ed6a04e2-e34c-4a70-89da-b1ec34d2fb5f\", \"text\": null, \"words\": null, \"utterances\": null, \"confidence\": null, \"audio_duration\": null, \"punctuate\": true, \"format_text\": true, \"dual_channel\": false, \"webhook_url\": null, \"webhook_status_code\": null, \"webhook_auth\": false, \"webhook_auth_header_name\": null, \"speed_boost\": false, \"auto_highlights_result\": null, \"auto_highlights\": false, \"audio_start_from\": null, \"audio_end_at\": null, \"word_boost\": [], \"boost_param\": null, \"prompt\": null, \"keyterms_prompt\": null, \"filter_profanity\": false, \"redact_pii\": false, \"redact_pii_audio\": false, \"redact_pii_audio_quality\": null, \"redact_pii_audio_options\": null, \"redact_pii_policies\": null, \"redact_pii_sub\": null, \"redact_static_entities\": null, \"speaker_labels\": true, \"speaker_options\": {\"min_speakers_expected\": null, \"max_speakers_expected\": null, \"advanced_speaker_segmentation\": true}, \"content_safety\": false, \"iab_categories\": false, \"content_safety_labels\": {}, \"iab_categories_result\": {}, \"language_detection\": true, \"language_detection_options\": {\"expected_languages\": null, \"fallback_language\": null, \"code_switching\": null, \"code_switching_confidence_threshold\": null, \"on_low_language_confidence\": null, \"on_no_speech_detected\": \"fallback\", \"swiss_german\": null, \"language_detection_model\": null, \"localization\": null}, \"language_detection_results\": null, \"language_confidence_threshold\": null, \"language_confidence\": null, \"custom_spelling\": null, \"throttled\": false, \"auto_chapters\": false, \"summarization\": false, \"summary_type\": null, \"summary_model\": null, \"custom_topics\": false, \"topics\": [], \"speech_threshold\": null, \"speech_model\": null, \"speech_models\": [\"universal-3-5-pro\"], \"speech_model_used\": null, \"temperature\": null, \"remove_audio_tags\": \"all\", \"chapters\": null, \"disfluencies\": false, \"entity_detection\": false, \"sentiment_analysis\": false, \"sentiment_analysis_results\": null, \"entities\": null, \"speakers_expected\": null, \"summary\": null, \"custom_topics_results\": null, \"is_deleted\": null, \"multichannel\": null, \"project_id\": 2040069, \"token_id\": 2077829}"
+      }
+    },
+    {
+      "transport": "http",
+      "request": {
+        "method": "GET",
+        "url": "https://api.eu.assemblyai.com/v2/transcript/a1de21ad-a178-4c28-b8f0-4b0334b72126",
+        "headers": {},
+        "body": ""
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "application/json; charset=UTF-8"
+        },
+        "body": "{\"id\": \"a1de21ad-a178-4c28-b8f0-4b0334b72126\", \"language_model\": \"assemblyai_default\", \"acoustic_model\": \"assemblyai_default\", \"language_code\": null, \"speech_understanding\": null, \"translated_texts\": null, \"status\": \"processing\", \"audio_url\": \"https://cdn.assemblyai.com/upload/f057ab4f09d5f28527be4b830b3d92f8bf6a97e5f854837cf22efbf2d09b5182/ed6a04e2-e34c-4a70-89da-b1ec34d2fb5f\", \"text\": null, \"words\": null, \"utterances\": null, \"confidence\": null, \"audio_duration\": null, \"punctuate\": true, \"format_text\": true, \"dual_channel\": false, \"webhook_url\": null, \"webhook_status_code\": null, \"webhook_auth\": false, \"webhook_auth_header_name\": null, \"speed_boost\": false, \"auto_highlights_result\": null, \"auto_highlights\": false, \"audio_start_from\": null, \"audio_end_at\": null, \"word_boost\": [], \"boost_param\": null, \"prompt\": null, \"keyterms_prompt\": [], \"filter_profanity\": false, \"redact_pii\": false, \"redact_pii_audio\": false, \"redact_pii_audio_quality\": null, \"redact_pii_audio_options\": null, \"redact_pii_policies\": null, \"redact_pii_sub\": null, \"redact_static_entities\": null, \"speaker_labels\": true, \"speaker_options\": {\"min_speakers_expected\": null, \"max_speakers_expected\": null, \"advanced_speaker_segmentation\": true}, \"content_safety\": false, \"iab_categories\": false, \"content_safety_labels\": {}, \"iab_categories_result\": {}, \"language_detection\": true, \"language_detection_options\": {\"expected_languages\": null, \"fallback_language\": null, \"code_switching\": null, \"code_switching_confidence_threshold\": null, \"on_low_language_confidence\": null, \"on_no_speech_detected\": \"fallback\", \"swiss_german\": null, \"language_detection_model\": null, \"localization\": null}, \"language_detection_results\": null, \"language_confidence_threshold\": null, \"language_confidence\": null, \"custom_spelling\": null, \"throttled\": false, \"auto_chapters\": false, \"summarization\": false, \"summary_type\": null, \"summary_model\": null, \"custom_topics\": false, \"topics\": [], \"speech_threshold\": null, \"speech_model\": null, \"speech_models\": [\"universal-3-5-pro\"], \"speech_model_used\": null, \"temperature\": null, \"remove_audio_tags\": \"all\", \"chapters\": null, \"disfluencies\": false, \"entity_detection\": false, \"sentiment_analysis\": false, \"sentiment_analysis_results\": null, \"entities\": null, \"speakers_expected\": null, \"summary\": null, \"custom_topics_results\": null, \"is_deleted\": null, \"multichannel\": null, \"project_id\": 2040069, \"token_id\": 2077829}"
+      }
+    },
+    {
+      "transport": "http",
+      "request": {
+        "method": "GET",
+        "url": "https://api.eu.assemblyai.com/v2/transcript/a1de21ad-a178-4c28-b8f0-4b0334b72126",
+        "headers": {},
+        "body": ""
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "application/json; charset=UTF-8"
+        },
+        "body": "{\"id\": \"a1de21ad-a178-4c28-b8f0-4b0334b72126\", \"language_model\": \"assemblyai_default\", \"acoustic_model\": \"assemblyai_default\", \"language_code\": \"en\", \"speech_understanding\": null, \"translated_texts\": null, \"status\": \"completed\", \"audio_url\": \"https://cdn.assemblyai.com/upload/f057ab4f09d5f28527be4b830b3d92f8bf6a97e5f854837cf22efbf2d09b5182/ed6a04e2-e34c-4a70-89da-b1ec34d2fb5f\", \"text\": \"Hello from Open Code.\", \"words\": [{\"text\": \"Hello\", \"start\": 445, \"end\": 633, \"confidence\": 0.8883743, \"speaker\": \"A\"}, {\"text\": \"from\", \"start\": 719, \"end\": 890, \"confidence\": 0.9978282, \"speaker\": \"A\"}, {\"text\": \"Open\", \"start\": 959, \"end\": 1147, \"confidence\": 0.9486656, \"speaker\": \"A\"}, {\"text\": \"Code.\", \"start\": 1215, \"end\": 1455, \"confidence\": 0.5913869, \"speaker\": \"A\"}], \"utterances\": [{\"speaker\": \"A\", \"text\": \"Hello from Open Code.\", \"confidence\": 0.85656375, \"start\": 445, \"end\": 1455, \"words\": [{\"text\": \"Hello\", \"start\": 445, \"end\": 633, \"confidence\": 0.8883743, \"speaker\": \"A\"}, {\"text\": \"from\", \"start\": 719, \"end\": 890, \"confidence\": 0.9978282, \"speaker\": \"A\"}, {\"text\": \"Open\", \"start\": 959, \"end\": 1147, \"confidence\": 0.9486656, \"speaker\": \"A\"}, {\"text\": \"Code.\", \"start\": 1215, \"end\": 1455, \"confidence\": 0.5913869, \"speaker\": \"A\"}]}], \"confidence\": 0.85656375, \"audio_duration\": 2, \"punctuate\": true, \"format_text\": true, \"dual_channel\": false, \"webhook_url\": null, \"webhook_status_code\": null, \"webhook_auth\": false, \"webhook_auth_header_name\": null, \"speed_boost\": false, \"auto_highlights_result\": null, \"auto_highlights\": false, \"audio_start_from\": null, \"audio_end_at\": null, \"word_boost\": [], \"boost_param\": null, \"prompt\": null, \"keyterms_prompt\": [], \"filter_profanity\": false, \"redact_pii\": false, \"redact_pii_audio\": false, \"redact_pii_audio_quality\": null, \"redact_pii_audio_options\": null, \"redact_pii_policies\": null, \"redact_pii_sub\": null, \"redact_static_entities\": null, \"speaker_labels\": true, \"speaker_options\": {\"min_speakers_expected\": null, \"max_speakers_expected\": null, \"advanced_speaker_segmentation\": true}, \"content_safety\": false, \"iab_categories\": false, \"content_safety_labels\": {\"status\": \"unavailable\", \"results\": [], \"summary\": {}}, \"iab_categories_result\": {\"status\": \"unavailable\", \"results\": [], \"summary\": {}}, \"language_detection\": true, \"language_detection_options\": {\"expected_languages\": null, \"fallback_language\": null, \"code_switching\": null, \"code_switching_confidence_threshold\": null, \"on_low_language_confidence\": null, \"on_no_speech_detected\": \"fallback\", \"swiss_german\": null, \"language_detection_model\": null, \"localization\": null}, \"language_detection_results\": null, \"language_confidence_threshold\": null, \"language_confidence\": 0.9934, \"custom_spelling\": null, \"throttled\": false, \"auto_chapters\": false, \"summarization\": false, \"summary_type\": null, \"summary_model\": null, \"custom_topics\": false, \"topics\": [], \"speech_threshold\": null, \"speech_model\": null, \"speech_models\": [\"universal-3-5-pro\"], \"speech_model_used\": \"universal-3-5-pro\", \"temperature\": null, \"remove_audio_tags\": \"all\", \"chapters\": null, \"disfluencies\": false, \"entity_detection\": false, \"sentiment_analysis\": false, \"sentiment_analysis_results\": null, \"entities\": null, \"speakers_expected\": null, \"summary\": null, \"custom_topics_results\": null, \"is_deleted\": null, \"multichannel\": null, \"project_id\": 2040069, \"token_id\": 2077829}"
+      }
+    },
+    {
+      "transport": "http",
+      "request": {
+        "method": "GET",
+        "url": "https://api.eu.assemblyai.com/v2/transcript/a1de21ad-a178-4c28-b8f0-4b0334b72126",
+        "headers": {},
+        "body": ""
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "application/json; charset=UTF-8"
+        },
+        "body": "{\"id\": \"a1de21ad-a178-4c28-b8f0-4b0334b72126\", \"language_model\": \"assemblyai_default\", \"acoustic_model\": \"assemblyai_default\", \"language_code\": \"en\", \"speech_understanding\": null, \"translated_texts\": null, \"status\": \"completed\", \"audio_url\": \"https://cdn.assemblyai.com/upload/f057ab4f09d5f28527be4b830b3d92f8bf6a97e5f854837cf22efbf2d09b5182/ed6a04e2-e34c-4a70-89da-b1ec34d2fb5f\", \"text\": \"Hello from Open Code.\", \"words\": [{\"text\": \"Hello\", \"start\": 445, \"end\": 633, \"confidence\": 0.8883743, \"speaker\": \"A\"}, {\"text\": \"from\", \"start\": 719, \"end\": 890, \"confidence\": 0.9978282, \"speaker\": \"A\"}, {\"text\": \"Open\", \"start\": 959, \"end\": 1147, \"confidence\": 0.9486656, \"speaker\": \"A\"}, {\"text\": \"Code.\", \"start\": 1215, \"end\": 1455, \"confidence\": 0.5913869, \"speaker\": \"A\"}], \"utterances\": [{\"speaker\": \"A\", \"text\": \"Hello from Open Code.\", \"confidence\": 0.85656375, \"start\": 445, \"end\": 1455, \"words\": [{\"text\": \"Hello\", \"start\": 445, \"end\": 633, \"confidence\": 0.8883743, \"speaker\": \"A\"}, {\"text\": \"from\", \"start\": 719, \"end\": 890, \"confidence\": 0.9978282, \"speaker\": \"A\"}, {\"text\": \"Open\", \"start\": 959, \"end\": 1147, \"confidence\": 0.9486656, \"speaker\": \"A\"}, {\"text\": \"Code.\", \"start\": 1215, \"end\": 1455, \"confidence\": 0.5913869, \"speaker\": \"A\"}]}], \"confidence\": 0.85656375, \"audio_duration\": 2, \"punctuate\": true, \"format_text\": true, \"dual_channel\": false, \"webhook_url\": null, \"webhook_status_code\": null, \"webhook_auth\": false, \"webhook_auth_header_name\": null, \"speed_boost\": false, \"auto_highlights_result\": null, \"auto_highlights\": false, \"audio_start_from\": null, \"audio_end_at\": null, \"word_boost\": [], \"boost_param\": null, \"prompt\": null, \"keyterms_prompt\": [], \"filter_profanity\": false, \"redact_pii\": false, \"redact_pii_audio\": false, \"redact_pii_audio_quality\": null, \"redact_pii_audio_options\": null, \"redact_pii_policies\": null, \"redact_pii_sub\": null, \"redact_static_entities\": null, \"speaker_labels\": true, \"speaker_options\": {\"min_speakers_expected\": null, \"max_speakers_expected\": null, \"advanced_speaker_segmentation\": true}, \"content_safety\": false, \"iab_categories\": false, \"content_safety_labels\": {\"status\": \"unavailable\", \"results\": [], \"summary\": {}}, \"iab_categories_result\": {\"status\": \"unavailable\", \"results\": [], \"summary\": {}}, \"language_detection\": true, \"language_detection_options\": {\"expected_languages\": null, \"fallback_language\": null, \"code_switching\": null, \"code_switching_confidence_threshold\": null, \"on_low_language_confidence\": null, \"on_no_speech_detected\": \"fallback\", \"swiss_german\": null, \"language_detection_model\": null, \"localization\": null}, \"language_detection_results\": null, \"language_confidence_threshold\": null, \"language_confidence\": 0.9934, \"custom_spelling\": null, \"throttled\": false, \"auto_chapters\": false, \"summarization\": false, \"summary_type\": null, \"summary_model\": null, \"custom_topics\": false, \"topics\": [], \"speech_threshold\": null, \"speech_model\": null, \"speech_models\": [\"universal-3-5-pro\"], \"speech_model_used\": \"universal-3-5-pro\", \"temperature\": null, \"remove_audio_tags\": \"all\", \"chapters\": null, \"disfluencies\": false, \"entity_detection\": false, \"sentiment_analysis\": false, \"sentiment_analysis_results\": null, \"entities\": null, \"speakers_expected\": null, \"summary\": null, \"custom_topics_results\": null, \"is_deleted\": null, \"multichannel\": null, \"project_id\": 2040069, \"token_id\": 2077829}"
+      }
+    }
+  ]
+}
@@ -0,0 +1,32 @@
+{
+  "version": 1,
+  "metadata": {
+    "tags": [
+      "prefix:deepgram-transcription",
+      "provider:deepgram",
+      "protocol:deepgram-transcription"
+    ],
+    "name": "deepgram-transcription/splits-utterances-at-speaker-changes",
+    "recordedAt": "2026-09-23T13:46:35.886Z"
+  },
+  "interactions": [
+    {
+      "transport": "http",
+      "request": {
+        "method": "POST",
+        "url": "https://api.deepgram.com/v1/listen?model=nova-3&smart_format=true&detect_language=true&diarize_model=latest&utterances=true",
+        "headers": {
+          "content-type": "audio/mpeg"
+        },
+        "body": "[audio]"
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "application/json"
+        },
+        "body": "{\"metadata\":{\"transaction_key\":\"deprecated\",\"request_id\":\"01a0ce84-a81d-7161-8505-c08d5e0c25bb\",\"sha256\":\"12e9ccb9d346c28b071c2fab7ce48556d0cc115af6850aacb338e7548807cb80\",\"created\":\"2026-09-23T13:46:35.566Z\",\"duration\":3.552625,\"channels\":1,\"models\":[\"2187e11a-3532-4498-b076-81fa530bdd49\"],\"model_info\":{\"2187e11a-3532-4498-b076-81fa530bdd49\":{\"name\":\"general-nova-3\",\"version\":\"2025-07-31.0\",\"arch\":\"nova-3\"}},\"diarize_info\":{\"model_uuid\":\"6946b038-8264-4505-acf7-9822c009fecb\",\"arch\":\"v2\"}},\"results\":{\"channels\":[{\"alternatives\":[{\"transcript\":\"Did the release ship? Yes. It shipped this morning.\",\"confidence\":1.0,\"words\":[{\"word\":\"did\",\"start\":0.08,\"end\":0.48,\"confidence\":1.0,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"Did\"},{\"word\":\"the\",\"start\":0.48,\"end\":0.64,\"confidence\":1.0,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"the\"},{\"word\":\"release\",\"start\":0.64,\"end\":0.96,\"confidence\":0.99902344,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"release\"},{\"word\":\"ship\",\"start\":0.96,\"end\":1.52,\"confidence\":1.0,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"ship?\"},{\"word\":\"yes\",\"start\":1.5999999,\"end\":2.32,\"confidence\":0.9980469,\"speaker\":1,\"speaker_confidence\":0.98,\"punctuated_word\":\"Yes.\"},{\"word\":\"it\",\"start\":2.32,\"end\":2.56,\"confidence\":1.0,\"speaker\":1,\"speaker_confidence\":0.98,\"punctuated_word\":\"It\"},{\"word\":\"shipped\",\"start\":2.56,\"end\":2.8799999,\"confidence\":0.9707031,\"speaker\":1,\"speaker_confidence\":0.98,\"punctuated_word\":\"shipped\"},{\"word\":\"this\",\"start\":2.8799999,\"end\":3.1999998,\"confidence\":1.0,\"speaker\":1,\"speaker_confidence\":0.98,\"punctuated_word\":\"this\"},{\"word\":\"morning\",\"start\":3.1999998,\"end\":3.52,\"confidence\":1.0,\"speaker\":1,\"speaker_confidence\":0.98,\"punctuated_word\":\"morning.\"}],\"paragraphs\":{\"transcript\":\"\\nSpeaker 0: Did the release ship?\\n\\nSpeaker 1: Yes. It shipped this morning.\",\"paragraphs\":[{\"sentences\":[{\"text\":\"Did the release ship?\",\"start\":0.08,\"end\":1.52}],\"speaker\":0,\"num_words\":4,\"start\":0.08,\"end\":1.52},{\"sentences\":[{\"text\":\"Yes.\",\"start\":1.5999999,\"end\":2.32},{\"text\":\"It shipped this morning.\",\"start\":2.32,\"end\":3.52}],\"speaker\":1,\"num_words\":5,\"start\":1.5999999,\"end\":3.52}]}}],\"detected_language\":\"en\",\"language_confidence\":0.917855}],\"utterances\":[{\"start\":0.08,\"end\":3.52,\"confidence\":0.99641925,\"channel\":0,\"transcript\":\"Did the release ship? Yes. It shipped this morning.\",\"words\":[{\"word\":\"did\",\"start\":0.08,\"end\":0.48,\"confidence\":1.0,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"Did\"},{\"word\":\"the\",\"start\":0.48,\"end\":0.64,\"confidence\":1.0,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"the\"},{\"word\":\"release\",\"start\":0.64,\"end\":0.96,\"confidence\":0.99902344,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"release\"},{\"word\":\"ship\",\"start\":0.96,\"end\":1.52,\"confidence\":1.0,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"ship?\"},{\"word\":\"yes\",\"start\":1.5999999,\"end\":2.32,\"confidence\":0.9980469,\"speaker\":1,\"speaker_confidence\":0.98,\"punctuated_word\":\"Yes.\"},{\"word\":\"it\",\"start\":2.32,\"end\":2.56,\"confidence\":1.0,\"speaker\":1,\"speaker_confidence\":0.98,\"punctuated_word\":\"It\"},{\"word\":\"shipped\",\"start\":2.56,\"end\":2.8799999,\"confidence\":0.9707031,\"speaker\":1,\"speaker_confidence\":0.98,\"punctuated_word\":\"shipped\"},{\"word\":\"this\",\"start\":2.8799999,\"end\":3.1999998,\"confidence\":1.0,\"speaker\":1,\"speaker_confidence\":0.98,\"punctuated_word\":\"this\"},{\"word\":\"morning\",\"start\":3.1999998,\"end\":3.52,\"confidence\":1.0,\"speaker\":1,\"speaker_confidence\":0.98,\"punctuated_word\":\"morning.\"}],\"speaker\":1,\"id\":\"67ba8055-e541-4089-abfa-0163558083cc\"}]}}"
+      }
+    }
+  ]
+}
@@ -0,0 +1,50 @@
+{
+  "version": 1,
+  "metadata": {
+    "tags": [
+      "prefix:deepgram-transcription",
+      "provider:deepgram",
+      "protocol:deepgram-transcription"
+    ],
+    "name": "deepgram-transcription/transcribes-raw-audio-with-speakers-and-word-timestamps",
+    "recordedAt": "2026-09-23T12:32:18.578Z"
+  },
+  "interactions": [
+    {
+      "transport": "http",
+      "request": {
+        "method": "POST",
+        "url": "https://api.deepgram.com/v1/listen?model=nova-3&smart_format=true&detect_language=true&diarize_model=latest&utterances=true",
+        "headers": {
+          "content-type": "audio/mpeg"
+        },
+        "body": "[audio]"
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "application/json"
+        },
+        "body": "{\"metadata\":{\"transaction_key\":\"deprecated\",\"request_id\":\"01a0ce40-a44f-73b0-9d99-cdfc4f184268\",\"sha256\":\"95495edaebf6b10b2877de8c7b9eb7372de315c105fa8e73ed656ef2ec28ca02\",\"created\":\"2026-09-23T12:32:18.036Z\",\"duration\":1.5411875,\"channels\":1,\"models\":[\"2187e11a-3532-4498-b076-81fa530bdd49\"],\"model_info\":{\"2187e11a-3532-4498-b076-81fa530bdd49\":{\"name\":\"general-nova-3\",\"version\":\"2025-07-31.0\",\"arch\":\"nova-3\"}},\"diarize_info\":{\"model_uuid\":\"6946b038-8264-4505-acf7-9822c009fecb\",\"arch\":\"v2\"}},\"results\":{\"channels\":[{\"alternatives\":[{\"transcript\":\"Hello from OpenCode.\",\"confidence\":0.96972656,\"words\":[{\"word\":\"hello\",\"start\":0.16,\"end\":0.64,\"confidence\":0.9892578,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"Hello\"},{\"word\":\"from\",\"start\":0.64,\"end\":0.88,\"confidence\":0.96972656,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"from\"},{\"word\":\"opencode\",\"start\":0.88,\"end\":1.4399999,\"confidence\":0.9472656,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"OpenCode.\"}],\"paragraphs\":{\"transcript\":\"\\nSpeaker 0: Hello from OpenCode.\",\"paragraphs\":[{\"sentences\":[{\"text\":\"Hello from OpenCode.\",\"start\":0.16,\"end\":1.4399999}],\"speaker\":0,\"num_words\":3,\"start\":0.16,\"end\":1.4399999}]}}],\"detected_language\":\"en\",\"language_confidence\":0.9998003}],\"utterances\":[{\"start\":0.16,\"end\":1.4399999,\"confidence\":0.96875,\"channel\":0,\"transcript\":\"Hello from OpenCode.\",\"words\":[{\"word\":\"hello\",\"start\":0.16,\"end\":0.64,\"confidence\":0.9892578,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"Hello\"},{\"word\":\"from\",\"start\":0.64,\"end\":0.88,\"confidence\":0.96972656,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"from\"},{\"word\":\"opencode\",\"start\":0.88,\"end\":1.4399999,\"confidence\":0.9472656,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"OpenCode.\"}],\"speaker\":0,\"id\":\"43d717d5-ca00-48c1-a39c-381be0b1dad2\"}]}}"
+      }
+    },
+    {
+      "transport": "http",
+      "request": {
+        "method": "POST",
+        "url": "https://api.deepgram.com/v1/listen?model=nova-3&smart_format=true&detect_language=true&diarize_model=latest&utterances=true",
+        "headers": {
+          "content-type": "audio/mpeg"
+        },
+        "body": "[audio]"
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "application/json"
+        },
+        "body": "{\"metadata\":{\"transaction_key\":\"deprecated\",\"request_id\":\"01a0ce40-a609-78c2-8a9d-806e3ff8db0b\",\"sha256\":\"95495edaebf6b10b2877de8c7b9eb7372de315c105fa8e73ed656ef2ec28ca02\",\"created\":\"2026-09-23T12:32:18.478Z\",\"duration\":1.5411875,\"channels\":1,\"models\":[\"2187e11a-3532-4498-b076-81fa530bdd49\"],\"model_info\":{\"2187e11a-3532-4498-b076-81fa530bdd49\":{\"name\":\"general-nova-3\",\"version\":\"2025-07-31.0\",\"arch\":\"nova-3\"}},\"diarize_info\":{\"model_uuid\":\"6946b038-8264-4505-acf7-9822c009fecb\",\"arch\":\"v2\"}},\"results\":{\"channels\":[{\"alternatives\":[{\"transcript\":\"Hello from OpenCode.\",\"confidence\":0.96972656,\"words\":[{\"word\":\"hello\",\"start\":0.16,\"end\":0.64,\"confidence\":0.9892578,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"Hello\"},{\"word\":\"from\",\"start\":0.64,\"end\":0.88,\"confidence\":0.96972656,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"from\"},{\"word\":\"opencode\",\"start\":0.88,\"end\":1.4399999,\"confidence\":0.9472656,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"OpenCode.\"}],\"paragraphs\":{\"transcript\":\"\\nSpeaker 0: Hello from OpenCode.\",\"paragraphs\":[{\"sentences\":[{\"text\":\"Hello from OpenCode.\",\"start\":0.16,\"end\":1.4399999}],\"speaker\":0,\"num_words\":3,\"start\":0.16,\"end\":1.4399999}]}}],\"detected_language\":\"en\",\"language_confidence\":0.9998003}],\"utterances\":[{\"start\":0.16,\"end\":1.4399999,\"confidence\":0.96875,\"channel\":0,\"transcript\":\"Hello from OpenCode.\",\"words\":[{\"word\":\"hello\",\"start\":0.16,\"end\":0.64,\"confidence\":0.9892578,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"Hello\"},{\"word\":\"from\",\"start\":0.64,\"end\":0.88,\"confidence\":0.96972656,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"from\"},{\"word\":\"opencode\",\"start\":0.88,\"end\":1.4399999,\"confidence\":0.9472656,\"speaker\":0,\"speaker_confidence\":0.98,\"punctuated_word\":\"OpenCode.\"}],\"speaker\":0,\"id\":\"6f46ea75-616b-4209-b644-aae3e8f1e3c0\"}]}}"
+      }
+    }
+  ]
+}
@@ -0,0 +1,32 @@
+{
+  "version": 1,
+  "metadata": {
+    "tags": [
+      "prefix:openai-transcription",
+      "provider:openai",
+      "protocol:openai-transcription"
+    ],
+    "name": "openai-transcription/diarizes-speakers-into-segments",
+    "recordedAt": "2026-09-23T12:32:01.274Z"
+  },
+  "interactions": [
+    {
+      "transport": "http",
+      "request": {
+        "method": "POST",
+        "url": "https://api.openai.com/v1/audio/transcriptions",
+        "headers": {
+          "content-type": "multipart/form-data; boundary=----WebKitFormBoundaryf476753a62264fa88f2d753a7e85c3a4"
+        },
+        "body": "--BOUNDARY\r\nContent-Disposition: form-data; name=\"file\"; filename=\"audio.mp3\"\r\nContent-Type: audio/mpeg\r\n\r\n[audio]\r\n--BOUNDARY\r\nContent-Disposition: form-data; name=\"model\"\r\n\r\ngpt-4o-transcribe-diarize\r\n--BOUNDARY\r\nContent-Disposition: form-data; name=\"response_format\"\r\n\r\ndiarized_json\r\n--BOUNDARY\r\nContent-Disposition: form-data; name=\"chunking_strategy\"\r\n\r\nauto\r\n--BOUNDARY--\r\n"
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "application/json"
+        },
+        "body": "{\"text\":\"Hello from OpenCode!\",\"task\":\"transcribe\",\"duration\":1.54125,\"segments\":[{\"type\":\"transcript.text.segment\",\"text\":\" Hello from OpenCode!\",\"speaker\":\"A\",\"start\":0.30000000000000004,\"end\":1.3499999999999999,\"id\":\"seg_0\"}],\"usage\":{\"type\":\"tokens\",\"total_tokens\":107,\"input_tokens\":15,\"input_token_details\":{\"text_tokens\":0,\"audio_tokens\":15},\"output_tokens\":92}}"
+      }
+    }
+  ]
+}
@@ -0,0 +1,32 @@
+{
+  "version": 1,
+  "metadata": {
+    "tags": [
+      "prefix:openai-transcription",
+      "provider:openai",
+      "protocol:openai-transcription"
+    ],
+    "name": "openai-transcription/returns-whisper-word-timestamps",
+    "recordedAt": "2026-09-23T12:32:02.972Z"
+  },
+  "interactions": [
+    {
+      "transport": "http",
+      "request": {
+        "method": "POST",
+        "url": "https://api.openai.com/v1/audio/transcriptions",
+        "headers": {
+          "content-type": "multipart/form-data; boundary=----WebKitFormBoundarydfe85869a7d6448994e94f2f41d9ce55"
+        },
+        "body": "--BOUNDARY\r\nContent-Disposition: form-data; name=\"file\"; filename=\"audio.mp3\"\r\nContent-Type: audio/mpeg\r\n\r\n[audio]\r\n--BOUNDARY\r\nContent-Disposition: form-data; name=\"model\"\r\n\r\nwhisper-1\r\n--BOUNDARY\r\nContent-Disposition: form-data; name=\"response_format\"\r\n\r\nverbose_json\r\n--BOUNDARY\r\nContent-Disposition: form-data; name=\"timestamp_granularities[]\"\r\n\r\nword\r\n--BOUNDARY--\r\n"
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "application/json"
+        },
+        "body": "{\"task\":\"transcribe\",\"language\":\"english\",\"duration\":1.5399999618530273,\"text\":\"Hello from OpenCode!\",\"words\":[{\"word\":\"Hello\",\"start\":0.0,\"end\":0.6200000047683716},{\"word\":\"from\",\"start\":0.6200000047683716,\"end\":0.8799999952316284},{\"word\":\"OpenCode\",\"start\":0.8799999952316284,\"end\":1.399999976158142}],\"usage\":{\"type\":\"duration\",\"seconds\":2}}"
+      }
+    }
+  ]
+}
@@ -0,0 +1,32 @@
+{
+  "version": 1,
+  "metadata": {
+    "tags": [
+      "prefix:openai-transcription",
+      "provider:openai",
+      "protocol:openai-transcription"
+    ],
+    "name": "openai-transcription/streams-text-deltas",
+    "recordedAt": "2026-09-23T12:31:59.149Z"
+  },
+  "interactions": [
+    {
+      "transport": "http",
+      "request": {
+        "method": "POST",
+        "url": "https://api.openai.com/v1/audio/transcriptions",
+        "headers": {
+          "content-type": "multipart/form-data; boundary=----WebKitFormBoundarya44da5204f47452ea1ef8dec145321dc"
+        },
+        "body": "--BOUNDARY\r\nContent-Disposition: form-data; name=\"file\"; filename=\"audio.mp3\"\r\nContent-Type: audio/mpeg\r\n\r\n[audio]\r\n--BOUNDARY\r\nContent-Disposition: form-data; name=\"model\"\r\n\r\ngpt-4o-mini-transcribe\r\n--BOUNDARY\r\nContent-Disposition: form-data; name=\"stream\"\r\n\r\ntrue\r\n--BOUNDARY--\r\n"
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "text/event-stream; charset=utf-8"
+        },
+        "body": "data: {\"type\":\"transcript.text.delta\",\"delta\":\"Hello\"}\r\n\r\ndata: {\"type\":\"transcript.text.delta\",\"delta\":\" from\"}\r\n\r\ndata: {\"type\":\"transcript.text.delta\",\"delta\":\" Open\"}\r\n\r\ndata: {\"type\":\"transcript.text.delta\",\"delta\":\"Code\"}\r\n\r\ndata: {\"type\":\"transcript.text.delta\",\"delta\":\".\"}\r\n\r\ndata: {\"type\":\"transcript.text.done\",\"text\":\"Hello from OpenCode.\",\"usage\":{\"type\":\"tokens\",\"total_tokens\":22,\"input_tokens\":15,\"input_token_details\":{\"text_tokens\":0,\"audio_tokens\":15},\"output_tokens\":7}}\r\n\r\ndata: [DONE]\r\n\r\n"
+      }
+    }
+  ]
+}
@@ -0,0 +1,32 @@
+{
+  "version": 1,
+  "metadata": {
+    "tags": [
+      "prefix:openai-transcription",
+      "provider:openai",
+      "protocol:openai-transcription"
+    ],
+    "name": "openai-transcription/transcribes-with-a-language-hint",
+    "recordedAt": "2026-09-23T12:31:56.657Z"
+  },
+  "interactions": [
+    {
+      "transport": "http",
+      "request": {
+        "method": "POST",
+        "url": "https://api.openai.com/v1/audio/transcriptions",
+        "headers": {
+          "content-type": "multipart/form-data; boundary=----WebKitFormBoundary484bf529b0064c43ac4a965c37433c9e"
+        },
+        "body": "--BOUNDARY\r\nContent-Disposition: form-data; name=\"file\"; filename=\"audio.mp3\"\r\nContent-Type: audio/mpeg\r\n\r\n[audio]\r\n--BOUNDARY\r\nContent-Disposition: form-data; name=\"model\"\r\n\r\ngpt-transcribe\r\n--BOUNDARY\r\nContent-Disposition: form-data; name=\"languages[]\"\r\n\r\nen\r\n--BOUNDARY--\r\n"
+      },
+      "response": {
+        "status": 200,
+        "headers": {
+          "content-type": "application/json"
+        },
+        "body": "{\"text\":\"Hello from OpenCode.\",\"languages\":[{\"code\":\"en\"}],\"usage\":{\"type\":\"duration\",\"seconds\":2}}"
+      }
+    }
+  ]
+}
@@ -23,15 +23,14 @@ const scriptedRoute = (statuses: ReadonlyArray<GenerationStatus>, result: string
      progress: count / statuses.length,
    })
    const route: GenerationRoute<string> = {
-      status: () => Ref.updateAndGet(polls, (count) => count + 1).pipe(Effect.map(snapshot)),
-      result: (token) =>
-        Effect.gen(function* () {
-          const count = yield* Ref.get(polls)
-          const status = snapshot(count).status
-          if (status === "completed") return `${result}:${String(token)}`
-          return yield* new AIError({ reason: new InvalidProviderOutputError({ message: `Generation ended ${status}` }) })
-        }),
-      cancel: () => Ref.set(cancelled, true),
+      status: Ref.updateAndGet(polls, (count) => count + 1).pipe(Effect.map(snapshot)),
+      result: Effect.gen(function* () {
+        const count = yield* Ref.get(polls)
+        const status = snapshot(count).status
+        if (status === "completed") return result
+        return yield* new AIError({ reason: new InvalidProviderOutputError({ message: `Generation ended ${status}` }) })
+      }),
+      cancel: Ref.set(cancelled, true),
    }
    return { route, polls, cancelled }
  })
@@ -47,7 +46,7 @@ describe("Generation", () => {
      yield* TestClock.adjust("3 seconds")
      const result = yield* Fiber.join(fiber)

-      expect(result).toBe("done:[object Object]")
+      expect(result).toBe("done")
      expect(yield* Ref.get(scripted.polls)).toBe(3)
    }),
  )
@@ -57,7 +56,7 @@ describe("Generation", () => {
      const scripted = yield* scriptedRoute(["completed"], "done")
      yield* Ref.set(scripted.polls, 1)
      const generation = new Generation(scripted.route, "t", { id: "gen_1", status: "completed" })
-      expect(yield* generation.await()).toBe("done:t")
+      expect(yield* generation.await()).toBe("done")
      expect(yield* Ref.get(scripted.polls)).toBe(1)
    }),
  )
@@ -80,6 +79,24 @@ describe("Generation", () => {
    }),
  )

+  it.effect("fails an event stream at the deadline even when a status poll hangs", () =>
+    Effect.gen(function* () {
+      const route: GenerationRoute<string> = {
+        status: Effect.never,
+        result: Effect.succeed("never"),
+      }
+      const generation = new Generation(route, "t", { id: "gen_1", status: "queued" })
+
+      const fiber = yield* Effect.forkChild(
+        generation.events({ poll: { interval: "1 second", timeout: "5 seconds" } }).pipe(Stream.runCollect, Effect.flip),
+      )
+      yield* TestClock.adjust("6 seconds")
+      const error = yield* Fiber.join(fiber)
+
+      expect(error.reason._tag).toBe("Timeout")
+    }),
+  )
+
  it.effect("surfaces the route failure body for failed generations", () =>
    Effect.gen(function* () {
      const scripted = yield* scriptedRoute(["running", "failed"], "unused")
@@ -99,7 +116,9 @@ describe("Generation", () => {
      const scripted = yield* scriptedRoute(["queued", "running", "completed"], "done")
      const generation = new Generation(scripted.route, "t", { id: "gen_1", status: "queued" })

-      const fiber = yield* Effect.forkChild(generation.events({ poll: { interval: "1 second" } }).pipe(Stream.runCollect))
+      const fiber = yield* Effect.forkChild(
+        generation.events({ poll: { interval: "1 second" } }).pipe(Stream.runCollect),
+      )
      yield* TestClock.adjust("3 seconds")
      const events = Array.from(yield* Fiber.join(fiber))

@@ -63,6 +63,13 @@ describe("Media", () => {
      expect(Media.detectMediaType(WEBP)).toBe("image/webp")
      expect(Media.detectMediaType(PDF)).toBe("application/pdf")
      expect(Media.detectMediaType(Uint8Array.from([1, 2, 3]))).toBeUndefined()
+      expect(Media.detectMediaType(new TextEncoder().encode("ID3\x04"))).toBe("audio/mpeg")
+      expect(Media.detectMediaType(Uint8Array.from([0xff, 0xfb, 0x90, 0x64]))).toBe("audio/mpeg")
+      expect(Media.detectMediaType(Uint8Array.from([0xff, 0xf3, 0x90, 0x64]))).toBe("audio/mpeg")
+      expect(Media.detectMediaType(Uint8Array.from([0xff, 0xf1, 0x50, 0x80]))).toBe("audio/aac")
+      expect(Media.detectMediaType(new TextEncoder().encode("RIFF\0\0\0\0WAVEfmt "))).toBe("audio/wav")
+      expect(Media.detectMediaType(new TextEncoder().encode("OggS\0\x02"))).toBe("audio/ogg")
+      expect(Media.detectMediaType(new TextEncoder().encode("fLaC\0\0\0\x22"))).toBe("audio/flac")
      expect(Media.bytes(PNG).mediaType).toBe("image/png")
      expect(Media.bytes(Uint8Array.from([1, 2, 3])).mediaType).toBe("application/octet-stream")
      expect(Media.bytes(Uint8Array.from([1, 2, 3]), "image/x-custom").mediaType).toBe("image/x-custom")
@@ -1,10 +1,10 @@
 import { describe, expect, test } from "bun:test"
 import { Effect, Layer } from "effect"
 import { HttpClientRequest } from "effect/unstable/http"
-import { AIError, LLMEvent, Media } from "../src/index.js"
+import { AIError, LLMEvent, Media, SpeechEvent, TranscriptionEvent } from "../src/index.js"
 import { RequestExecutor } from "../src/route.js"
 import { AI } from "../src/promise.js"
-import { OpenAI } from "../src/providers.js"
+import { AssemblyAI, OpenAI, Runway } from "../src/providers.js"
 import { handlerLayer } from "./lib/http.js"
 import { sseEvents } from "./lib/sse.js"

@@ -16,7 +16,11 @@ const chatBody = sseEvents(
  { choices: [{ delta: {}, finish_reason: "stop" }] },
 )

-/** Executor layer that answers chat completions with SSE text and image generations with one base64 PNG. */
+/**
+ * Executor layer that answers chat completions with SSE text, image generations with one base64 PNG, Runway video
+ * tasks with a queued submission that succeeds on the second poll, speech with raw audio or SSE audio deltas, OpenAI
+ * transcription with JSON or SSE text deltas, and AssemblyAI transcripts that complete on the first poll.
+ */
 const executor = (seen: Array<string>) =>
  RequestExecutor.layer.pipe(
    Layer.provide(
@@ -30,6 +34,51 @@ const executor = (seen: Array<string>) =>
            })
          if (web.url.endsWith("/chat/completions"))
            return input.respond(chatBody, { headers: { "content-type": "text/event-stream" } })
+          if (web.url.endsWith("/audio/speech"))
+            return JSON.parse(input.text).stream_format === "sse"
+              ? input.respond(
+                  sseEvents(
+                    { type: "speech.audio.delta", audio: "AQI=" },
+                    { type: "speech.audio.delta", audio: "Aw==" },
+                    { type: "speech.audio.done", usage: { input_tokens: 4, output_tokens: 8, total_tokens: 12 } },
+                  ),
+                  { headers: { "content-type": "text/event-stream" } },
+                )
+              : input.respond(Uint8Array.from([1, 2, 3]), { headers: { "content-type": "audio/pcm" } })
+          if (web.url.endsWith("/audio/transcriptions"))
+            return input.text.includes('name="stream"')
+              ? input.respond(
+                  sseEvents(
+                    { type: "transcript.text.delta", delta: "Hello" },
+                    { type: "transcript.text.delta", delta: " there." },
+                    { type: "transcript.text.done", text: "Hello there." },
+                  ),
+                  { headers: { "content-type": "text/event-stream" } },
+                )
+              : input.respond(JSON.stringify({ text: "Hello there." }), {
+                  headers: { "content-type": "application/json" },
+                })
+          if (web.url.endsWith("/v2/transcript"))
+            return input.respond(JSON.stringify({ id: "tr_1", status: "queued" }), {
+              headers: { "content-type": "application/json" },
+            })
+          if (web.url.endsWith("/v2/transcript/tr_1"))
+            return input.respond(JSON.stringify({ id: "tr_1", status: "completed", text: "Hello there." }), {
+              headers: { "content-type": "application/json" },
+            })
+          if (web.url.endsWith("/text_to_video"))
+            return input.respond(JSON.stringify({ id: "task_1" }), { headers: { "content-type": "application/json" } })
+          if (web.url.endsWith("/tasks/task_1")) {
+            const polls = seen.filter((url) => url.endsWith("/tasks/task_1")).length
+            return input.respond(
+              JSON.stringify(
+                polls === 1
+                  ? { status: "RUNNING", progress: 0.5 }
+                  : { status: "SUCCEEDED", output: ["https://runway.test/out.mp4"], cost: { credits: 5 } },
+              ),
+              { headers: { "content-type": "application/json" } },
+            )
+          }
          return input.respond(JSON.stringify({ error: { message: "not found" } }), {
            status: 404,
            headers: { "content-type": "application/json" },
@@ -73,6 +122,81 @@ describe("AI promise client", () => {
    await ai.dispose()
  })

+  test("starts, awaits, and resumes video generations over the same runtime", async () => {
+    const seen: Array<string> = []
+    const ai = AI.make({ layer: executor(seen) })
+    const model = Runway.configure({ apiKey: "test", baseURL: "https://runway.test/v1" }).video("gen4.5")
+
+    const generation = await ai.video.start({ model, prompt: "A kite" })
+    expect(generation.id).toBe("task_1")
+    expect(generation.status).toBe("queued")
+    expect(generation.token).toEqual({ taskID: "task_1" })
+
+    const refreshed = await generation.refresh()
+    expect(refreshed.status).toBe("running")
+    expect(refreshed.progress).toBe(0.5)
+
+    const response = await refreshed.await({ poll: { interval: 10 } })
+    expect(response.video).toBeInstanceOf(Media.Asset)
+    expect(response.video.source).toMatchObject({ type: "url", url: "https://runway.test/out.mp4" })
+    expect(response.usage).toEqual({ type: "credits", credits: 5 })
+
+    const resumed = await ai.video.resume(model, JSON.parse(JSON.stringify(generation.token)))
+    expect(resumed.status).toBe("completed")
+    expect((await resumed.await()).videos).toHaveLength(1)
+
+    const events: Array<string> = []
+    for await (const event of ai.video.stream({ model, prompt: "A kite" }, { poll: { interval: 10 } })) {
+      events.push(event.type)
+    }
+    expect(events).toEqual(["video", "finish"])
+
+    expect(seen[0]).toBe("https://runway.test/v1/text_to_video")
+    expect(seen.filter((url) => url.endsWith("/tasks/task_1")).length).toBeGreaterThanOrEqual(5)
+    await ai.dispose()
+  })
+
+  test("generates, streams, and starts transcriptions over the same runtime", async () => {
+    const ai = AI.make({ layer: executor([]) })
+    const audio = Media.url("https://audio.test/hello.mp3")
+    const bytes = Media.bytes(Uint8Array.from([0x49, 0x44, 0x33]), "audio/mpeg")
+
+    expect(
+      (await ai.transcription.generate({ model: openai.transcription("gpt-4o-mini-transcribe"), audio: bytes })).text,
+    ).toBe("Hello there.")
+
+    const deltas: Array<string> = []
+    for await (const event of ai.transcription.stream({
+      model: openai.transcription("gpt-4o-mini-transcribe"),
+      audio: bytes,
+    }))
+      if (TranscriptionEvent.is.textDelta(event)) deltas.push(event.delta)
+    expect(deltas).toEqual(["Hello", " there."])
+
+    const model = AssemblyAI.configure({ apiKey: "test", baseURL: "https://assemblyai.test" }).transcription(
+      "universal-2",
+    )
+    const generation = await ai.transcription.start({ model, audio })
+    expect(generation.token).toEqual({ transcriptID: "tr_1" })
+    expect((await generation.await({ poll: { interval: 10 } })).text).toBe("Hello there.")
+    await ai.dispose()
+  })
+
+  test("generates and streams speech over the same runtime", async () => {
+    const ai = AI.make({ layer: executor([]) })
+    const model = openai.speech("gpt-4o-mini-tts")
+
+    const response = await ai.speech.generate({ model, text: "Hello", voice: "coral", format: "pcm" })
+    expect(await ai.run(response.audio.bytes())).toEqual(Uint8Array.from([1, 2, 3]))
+
+    const events: Array<SpeechEvent> = []
+    for await (const event of ai.speech.stream({ model, text: "Hello", voice: "coral" })) events.push(event)
+    expect(events.map((event) => event.type)).toEqual(["audio-delta", "audio-delta", "finish"])
+    const finish = events.find(SpeechEvent.is.finish)
+    expect(await ai.run(finish!.audio.bytes())).toEqual(Uint8Array.from([1, 2, 3]))
+    await ai.dispose()
+  })
+
  test("rethrows AIError unchanged and honors abort signals", async () => {
    const ai = AI.make({ layer: executor([]) })

@@ -97,6 +221,8 @@ describe("AI promise client", () => {
  test("the default client is created lazily and can be disposed", async () => {
    expect(typeof AI.ai.llm.generate).toBe("function")
    expect(typeof AI.ai.image.generate).toBe("function")
+    expect(typeof AI.ai.video.start).toBe("function")
+    expect(typeof AI.ai.speech.generate).toBe("function")
    await AI.ai.dispose()
  })
 })
@@ -0,0 +1,39 @@
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { Transcription } from "../../src/index.js"
+import { AssemblyAI } from "../../src/providers.js"
+import { recordedTests } from "../recorded-test.js"
+import { TRANSCRIPT, audio, audioRecording } from "./transcription-recording.js"
+import { queuedPoll } from "./queued-recording.js"
+
+// Recorded against the EU region: the US endpoint was unreachable from the recording network. Same API and shapes.
+const model = AssemblyAI.configure({
+  apiKey: process.env.ASSEMBLYAI_API_KEY ?? "fixture",
+  baseURL: "https://api.eu.assemblyai.com",
+}).transcription("universal-3-5-pro")
+
+const recorded = recordedTests({
+  prefix: "assemblyai-transcription",
+  provider: "assemblyai",
+  protocol: "assemblyai-transcription",
+  requires: ["ASSEMBLYAI_API_KEY"],
+  options: audioRecording,
+})
+
+describe("AssemblyAI Transcription recorded", () => {
+  recorded.effect(
+    "uploads, submits, and polls a transcript",
+    () =>
+      Effect.gen(function* () {
+        const generation = yield* Transcription.start({ model, audio: yield* audio, diarize: true })
+        const response = yield* generation.await({ poll: queuedPoll })
+
+        expect(response.text).toMatch(TRANSCRIPT)
+        expect(response.words?.[0]).toMatchObject({ text: "Hello", speaker: "A" })
+        expect(response.words?.every((word) => word.startSeconds < 5 && word.endSeconds < 5)).toBe(true)
+        expect(response.segments?.[0]?.speaker).toBe("A")
+        expect(response.usage).toEqual({ type: "seconds", seconds: expect.any(Number) })
+      }),
+    { timeout: 15 * 60 * 1000 },
+  )
+})
@@ -0,0 +1,47 @@
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { Speech, SpeechEvent } from "../../src/index.js"
+import { Cartesia } from "../../src/providers.js"
+import { recordedTests } from "../recorded-test.js"
+import { TEXT, collectSpeech } from "./speech-recording.js"
+
+const model = Cartesia.configure({
+  apiKey: process.env.CARTESIA_API_KEY ?? "fixture",
+}).speech("sonic-3")
+
+// "Katie", the voice from Cartesia's quickstart.
+const voice = "f786b574-daa5-4673-aa0c-cbe3e8534c02"
+
+const recorded = recordedTests({
+  prefix: "cartesia-speech",
+  provider: "cartesia",
+  protocol: "cartesia-speech",
+  requires: ["CARTESIA_API_KEY"],
+  options: { redact: { allowRequestHeaders: ["cartesia-version"] } },
+})
+
+describe("Cartesia Speech recorded", () => {
+  recorded.effect("generates speech", () =>
+    Effect.gen(function* () {
+      const response = yield* Speech.generate({ model, text: TEXT, voice, format: "mp3" })
+
+      expect(response.audio.mediaType).toBe("audio/mpeg")
+      expect(response.audio.info).toMatchObject({ format: "mp3", sampleRate: 44100 })
+      expect((yield* response.audio.bytes()).length).toBeGreaterThan(0)
+    }),
+  )
+
+  recorded.effect("streams speech with timestamps", () =>
+    Effect.gen(function* () {
+      const { events, finish } = yield* collectSpeech(
+        Speech.stream({ model, text: TEXT, voice, timestamps: true, providerOptions: { sampleRate: 24000 } }),
+      )
+      const words = events.filter(SpeechEvent.is.timestamps).flatMap((event) => event.items)
+
+      expect(finish.audio.mediaType).toBe("audio/pcm")
+      expect(finish.audio.info).toMatchObject({ encoding: "pcm_s16le", sampleRate: 24000, channels: 1 })
+      expect(words.length).toBeGreaterThan(0)
+      expect(words.every((item) => item.endSeconds >= item.startSeconds)).toBe(true)
+    }),
+  )
+})
@@ -0,0 +1,41 @@
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { Speech } from "../../src/index.js"
+import { Deepgram } from "../../src/providers.js"
+import { recordedTests } from "../recorded-test.js"
+import { TEXT, collectSpeech } from "./speech-recording.js"
+
+const model = Deepgram.configure({
+  apiKey: process.env.DEEPGRAM_API_KEY ?? "fixture",
+}).speech("aura-2-thalia-en")
+
+const recorded = recordedTests({
+  prefix: "deepgram-speech",
+  provider: "deepgram",
+  protocol: "deepgram-speech",
+  requires: ["DEEPGRAM_API_KEY"],
+  // Usage and the request id arrive only as response headers.
+  options: { redact: { allowResponseHeaders: ["dg-char-count", "dg-request-id", "dg-model-name"] } },
+})
+
+describe("Deepgram Speech recorded", () => {
+  recorded.effect("generates speech", () =>
+    Effect.gen(function* () {
+      const response = yield* Speech.generate({ model, text: TEXT })
+
+      expect(response.audio.mediaType).toBe("audio/mpeg")
+      expect(response.audio.info).toEqual({ format: "mp3" })
+      expect((yield* response.audio.bytes()).length).toBeGreaterThan(0)
+      expect(response.usage).toEqual({ type: "characters", characters: expect.any(Number) })
+    }),
+  )
+
+  recorded.effect("streams speech", () =>
+    Effect.gen(function* () {
+      const { finish } = yield* collectSpeech(Speech.stream({ model, text: TEXT, format: "pcm" }))
+
+      expect(finish.audio.info).toMatchObject({ format: "pcm", encoding: "pcm_s16le", channels: 1 })
+      expect(finish.usage).toEqual({ type: "characters", characters: expect.any(Number) })
+    }),
+  )
+})
@@ -0,0 +1,47 @@
+import { describe, expect } from "bun:test"
+import { Effect, Stream } from "effect"
+import { Transcription } from "../../src/index.js"
+import { Deepgram } from "../../src/providers.js"
+import { recordedTests } from "../recorded-test.js"
+import { TRANSCRIPT, audio, audioRecording, dialog } from "./transcription-recording.js"
+
+const model = Deepgram.configure({ apiKey: process.env.DEEPGRAM_API_KEY ?? "fixture" }).transcription("nova-3")
+
+const recorded = recordedTests({
+  prefix: "deepgram-transcription",
+  provider: "deepgram",
+  protocol: "deepgram-transcription",
+  requires: ["DEEPGRAM_API_KEY"],
+  options: audioRecording,
+})
+
+describe("Deepgram Transcription recorded", () => {
+  recorded.effect("transcribes raw audio with speakers and word timestamps", () =>
+    Effect.gen(function* () {
+      const request = Transcription.request({ model, audio: yield* audio, diarize: true, timestamps: "word" })
+      const response = yield* Transcription.generate(request)
+
+      expect(response.text).toMatch(TRANSCRIPT)
+      expect(response.words?.map((word) => word.text)).toEqual(["Hello", "from", "OpenCode."])
+      expect(response.words?.every((word) => word.speaker === "0" && word.confidence !== undefined)).toBe(true)
+      expect(response.segments).toEqual([
+        { text: response.text, startSeconds: expect.any(Number), endSeconds: expect.any(Number), speaker: "0" },
+      ])
+      expect(response.language).toBe("en")
+      expect(response.usage).toEqual({ type: "seconds", seconds: response.durationSeconds })
+
+      const events = Array.from(yield* Stream.runCollect(Transcription.stream(request)))
+      expect(events.map((event) => event.type)).toEqual(["finish"])
+    }),
+  )
+
+  recorded.effect("splits utterances at speaker changes", () =>
+    Effect.gen(function* () {
+      const response = yield* Transcription.generate({ model, audio: yield* dialog, diarize: true })
+
+      expect(response.segments?.map((segment) => segment.speaker)).toEqual(["0", "1"])
+      expect(response.segments?.[0].text).toMatch(/release ship\?$/)
+      expect(response.segments?.[1].text).toMatch(/shipped this morning\.$/)
+    }),
+  )
+})
@@ -0,0 +1,54 @@
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { Speech, SpeechEvent } from "../../src/index.js"
+import { ElevenLabs } from "../../src/providers.js"
+import { recordedTests } from "../recorded-test.js"
+import { TEXT, collectSpeech } from "./speech-recording.js"
+
+const model = ElevenLabs.configure({
+  apiKey: process.env.ELEVENLABS_API_KEY ?? "fixture",
+}).speech("eleven_flash_v2_5")
+
+// "George", an ElevenLabs premade voice.
+const voice = "JBFqnCBsd6RMkjVDRZzb"
+
+const recorded = recordedTests({
+  prefix: "elevenlabs-speech",
+  provider: "elevenlabs",
+  protocol: "elevenlabs-speech",
+  requires: ["ELEVENLABS_API_KEY"],
+  // Usage and the request id arrive only as response headers.
+  options: { redact: { allowResponseHeaders: ["character-cost", "request-id"] } },
+})
+
+describe("ElevenLabs Speech recorded", () => {
+  recorded.effect("generates speech", () =>
+    Effect.gen(function* () {
+      const response = yield* Speech.generate({ model, text: TEXT, voice })
+
+      expect(response.audio.mediaType).toBe("audio/mpeg")
+      expect(response.audio.info).toMatchObject({ format: "mp3", sampleRate: 44100 })
+      expect((yield* response.audio.bytes()).length).toBeGreaterThan(0)
+      expect(response.usage).toEqual({ type: "credits", credits: 3 })
+    }),
+  )
+
+  recorded.effect("streams speech with timestamps", () =>
+    Effect.gen(function* () {
+      const { events, finish } = yield* collectSpeech(
+        Speech.stream({ model, text: TEXT, voice, format: "pcm", timestamps: true }),
+      )
+      const timestamps = events.filter(SpeechEvent.is.timestamps).flatMap((event) => event.items)
+
+      expect(finish.audio.mediaType).toBe("audio/pcm")
+      expect(finish.audio.info).toMatchObject({ encoding: "pcm_s16le", sampleRate: 24000, channels: 1 })
+      expect(
+        timestamps
+          .map((item) => item.text)
+          .join("")
+          .trim(),
+      ).toBe(TEXT)
+      expect(timestamps.every((item) => item.endSeconds >= item.startSeconds)).toBe(true)
+    }),
+  )
+})
@@ -0,0 +1,41 @@
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { Video } from "../../src/index.js"
+import { Fal } from "../../src/providers.js"
+import { recordedTests } from "../recorded-test.js"
+import { queuedPoll } from "./queued-recording.js"
+
+const model = Fal.configure({
+  apiKey: process.env.FAL_KEY ?? "fixture",
+}).video("fal-ai/veo3.1/fast")
+
+const recorded = recordedTests({
+  prefix: "fal-video",
+  provider: "fal",
+  protocol: "fal-video",
+  requires: ["FAL_KEY"],
+})
+
+describe("fal Video recorded", () => {
+  recorded.effect(
+    "generates a video",
+    () =>
+      Effect.gen(function* () {
+        const response = yield* Video.generate(
+          {
+            model,
+            prompt: "A single red balloon drifting slowly upward against a clear blue sky.",
+            aspectRatio: "16:9",
+            resolution: "720p",
+            providerOptions: { duration: "4s" },
+          },
+          { poll: queuedPoll },
+        )
+
+        expect(response.videos).toHaveLength(1)
+        expect(response.video.source.type).toBe("url")
+        expect((yield* response.video.bytes()).length).toBeGreaterThan(0)
+      }),
+    { timeout: 15 * 60 * 1000 },
+  )
+})
@@ -0,0 +1,45 @@
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { Speech } from "../../src/index.js"
+import { Google } from "../../src/providers.js"
+import { recordedTests } from "../recorded-test.js"
+import { TEXT, collectSpeech } from "./speech-recording.js"
+
+const google = Google.configure({ apiKey: process.env.GOOGLE_GENERATIVE_AI_API_KEY ?? "fixture" })
+
+const recorded = recordedTests({
+  prefix: "google-speech",
+  provider: "google",
+  protocol: "google-speech",
+  requires: ["GOOGLE_GENERATIVE_AI_API_KEY"],
+})
+
+describe("Google Speech recorded", () => {
+  recorded.effect("generates speech", () =>
+    Effect.gen(function* () {
+      const response = yield* Speech.generate({
+        model: google.speech("gemini-2.5-flash-preview-tts"),
+        text: `Say: ${TEXT}`,
+        voice: "Kore",
+      })
+
+      expect(response.audio.mediaType).toMatch(/^audio\/l16/i)
+      expect(response.audio.info).toMatchObject({ encoding: "pcm_s16le", sampleRate: 24000, channels: 1 })
+      expect((yield* response.audio.bytes()).length).toBeGreaterThan(0)
+      expect(response.usage?.type).toBe("tokens")
+    }),
+  )
+
+  // Only Gemini 3.1 TTS and later support `streamGenerateContent`.
+  recorded.effect("streams speech", () =>
+    Effect.gen(function* () {
+      const { finish } = yield* collectSpeech(
+        Speech.stream({ model: google.speech("gemini-3.1-flash-tts-preview"), text: `Say: ${TEXT}`, voice: "Kore" }),
+      )
+
+      expect(finish.audio.mediaType).toMatch(/^audio\/l16/i)
+      expect(finish.audio.info).toMatchObject({ encoding: "pcm_s16le", sampleRate: 24000, channels: 1 })
+      expect(finish.usage?.type).toBe("tokens")
+    }),
+  )
+})
@@ -0,0 +1,54 @@
+import { describe, expect } from "bun:test"
+import { Effect, Stream } from "effect"
+import { Transcription, TranscriptionEvent } from "../../src/index.js"
+import { Google } from "../../src/providers.js"
+import { recordedTests } from "../recorded-test.js"
+import { TRANSCRIPT, audio, audioRecording, dialog } from "./transcription-recording.js"
+
+const model = Google.configure({ apiKey: process.env.GOOGLE_GENERATIVE_AI_API_KEY ?? "fixture" }).transcription(
+  "gemini-3.5-transcribe",
+)
+
+const recorded = recordedTests({
+  prefix: "google-transcription",
+  provider: "google",
+  protocol: "google-transcription",
+  requires: ["GOOGLE_GENERATIVE_AI_API_KEY"],
+  options: audioRecording,
+})
+
+describe("Google Transcription recorded", () => {
+  recorded.effect("transcribes with speakers and word timestamps", () =>
+    Effect.gen(function* () {
+      const response = yield* Transcription.generate({ model, audio: yield* audio, diarize: true, language: "en-US" })
+
+      expect(response.text).toMatch(TRANSCRIPT)
+      expect(response.words?.length).toBeGreaterThan(2)
+      expect(response.words?.every((word) => word.endSeconds > word.startSeconds)).toBe(true)
+      expect(response.segments).toEqual([
+        {
+          text: response.text,
+          startSeconds: response.words?.[0]?.startSeconds,
+          endSeconds: response.words?.at(-1)?.endSeconds,
+          speaker: expect.any(String),
+        },
+      ])
+      expect(response.usage?.type).toBe("tokens")
+    }),
+  )
+
+  recorded.effect("streams speaker turns", () =>
+    Effect.gen(function* () {
+      const events = Array.from(
+        yield* Stream.runCollect(Transcription.stream({ model, audio: yield* dialog, diarize: true })),
+      )
+      const finish = events.at(-1)
+      if (finish === undefined || !TranscriptionEvent.is.finish(finish)) throw new Error("Expected a finish event")
+
+      expect(finish.text).toMatch(/release ship\? Yes, it shipped this morning\.$/)
+      const speakers = events.filter(TranscriptionEvent.is.segment).map((event) => event.segment.speaker)
+      expect(speakers).toHaveLength(2)
+      expect(new Set(speakers).size).toBe(2)
+    }),
+  )
+})
@@ -0,0 +1,41 @@
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { Video } from "../../src/index.js"
+import { Google } from "../../src/providers.js"
+import { recordedTests } from "../recorded-test.js"
+import { queuedPoll } from "./queued-recording.js"
+
+const model = Google.configure({
+  apiKey: process.env.GOOGLE_GENERATIVE_AI_API_KEY ?? "fixture",
+}).video("veo-3.1-fast-generate-preview")
+
+const recorded = recordedTests({
+  prefix: "google-video",
+  provider: "google",
+  protocol: "google-video",
+  requires: ["GOOGLE_GENERATIVE_AI_API_KEY"],
+})
+
+describe("Google Veo recorded", () => {
+  recorded.effect(
+    "generates a video",
+    () =>
+      Effect.gen(function* () {
+        const response = yield* Video.generate(
+          {
+            model,
+            prompt: "A single red balloon drifting slowly upward against a clear blue sky.",
+            aspectRatio: "16:9",
+            durationSeconds: 4,
+          },
+          { poll: queuedPoll },
+        )
+
+        expect(response.videos).toHaveLength(1)
+        expect(response.video.source.type).toBe("url")
+        expect(response.video.headers?.["x-goog-api-key"]).toBeString()
+        expect((yield* response.video.bytes()).length).toBeGreaterThan(0)
+      }),
+    { timeout: 15 * 60 * 1000 },
+  )
+})
@@ -791,6 +791,89 @@ describe("OpenAI Chat route", () => {
    }),
  )

+  it.effect("lowers inline PDFs as file parts", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            Message.user([
+              { type: "text", text: "Summarize these." },
+              { type: "media", media: Media.base64("JVBERi0=", "application/pdf"), filename: "report.pdf" },
+              { type: "media", media: Media.fromDataUrl("data:application/pdf;base64,JVBERi0=") },
+            ]),
+          ],
+        }),
+      )
+      expect(prepared.body.messages).toEqual([
+        {
+          role: "user",
+          content: [
+            { type: "text", text: "Summarize these." },
+            { type: "file", file: { filename: "report.pdf", file_data: "data:application/pdf;base64,JVBERi0=" } },
+            { type: "file", file: { filename: "document.pdf", file_data: "data:application/pdf;base64,JVBERi0=" } },
+          ],
+        },
+      ])
+    }),
+  )
+
+  it.effect("moves PDFs from tool results into a follow-up user message", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            Message.user("Read the report."),
+            Message.assistant([ToolCallPart.make({ id: "call_pdf", name: "read", input: {} })]),
+            Message.tool({
+              id: "call_pdf",
+              name: "read",
+              resultType: "content",
+              result: [
+                { type: "text", text: "PDF read successfully" },
+                {
+                  type: "file",
+                  mime: "application/pdf",
+                  uri: "data:application/pdf;base64,JVBERi0=",
+                  name: "report.pdf",
+                },
+              ],
+            }),
+          ],
+        }),
+      )
+      expect(prepared.body.messages).toContainEqual({
+        role: "tool",
+        tool_call_id: "call_pdf",
+        content: "PDF read successfully",
+      })
+      expect(prepared.body.messages.at(-1)).toEqual({
+        role: "user",
+        content: [
+          { type: "file", file: { filename: "report.pdf", file_data: "data:application/pdf;base64,JVBERi0=" } },
+        ],
+      })
+    }),
+  )
+
+  it.effect("requires inline data for PDF files", () =>
+    Effect.gen(function* () {
+      const error = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            Message.user({
+              type: "media",
+              media: Media.url("https://example.com/report.pdf", { mediaType: "application/pdf" }),
+            }),
+          ],
+        }),
+      ).pipe(Effect.flip)
+      expect(error.message).toContain("OpenAI Chat requires inline media")
+    }),
+  )
+
  it.effect("prepares raw and data URL image media as vision input", () =>
    Effect.gen(function* () {
      const prepared = yield* compileRequest(
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Jack	e64e4d0fbb	docs: add Space Bunny to V2 console docs (#50940 )	2026-09-24 01:03:39 +08:00
opencode-agent[bot]andMrMushrooooom	7f39ae9ecb	fix(stats): attribute Hy4 preview traffic to Tencent (#50935 ) Co-authored-by: MrMushrooooom <19261047+MrMushrooooom@users.noreply.github.com>	2026-09-24 00:56:42 +08:00
James Long	b3b030856b	feat(tui): budget transcript mounting by rendered entries (#50936 )	2026-09-23 12:36:10 -04:00
Aiden Cline	a117ebb408	fix(client): include the detail in client error messages (#50929 )	2026-09-23 11:33:49 -05:00
Aiden Cline	f2d9ebb886	fix(ai): send PDFs as file parts on OpenAI Chat routes (#50933 )	2026-09-23 11:27:33 -05:00
James Long	0332a26be6	fix(tui): load history until the oldest group is complete (#50930 )	2026-09-23 11:58:44 -04:00
opencode-agent[bot]	bb7876dfa8	chore: update nix node_modules hashes	2026-09-23 15:48:10 +00:00
Shoubhit Dash	dbaa57a21b	feat(ai): add transcription across inline, streamed, and queued routes (#50910 )	2026-09-23 21:15:50 +05:30
James Long	b6bc55764c	refactor(tui): persist group expansion and exact scroll anchors (#48489 )	2026-09-23 11:33:19 -04:00
Aiden Cline	14a3311a61	feat(codemode): add Uint8Array callback methods and coerce indexOf start indexes (#50829 )	2026-09-23 10:27:40 -05:00
opencode-agent[bot]andnexxeln	dddb5eb96f	fix(util): update OpenTelemetry to patched core (#50867 ) Co-authored-by: nexxeln <95541290+nexxeln@users.noreply.github.com>	2026-09-23 20:57:16 +05:30
Shoubhit Dash	affa57e40f	fix(core): relax websocket idle timeout and honor chunkTimeout (#50914 )	2026-09-23 20:22:41 +05:30
opencode-agent[bot]	681ea07a92	chore(core): refresh bundled models.dev snapshot	2026-09-23 12:20:47 +00:00
Adam	6e89d9e2c3	feat(core): identify the client on Console requests (#50896 )	2026-09-23 07:09:19 -05:00
Shoubhit Dash	f526727178	feat(ai): add speech generation with streaming media routes (#50883 )	2026-09-23 17:35:23 +05:30
Luke Parker	9c8a63e852	fix(app): keep composer focus after attaching files (#50864 )	2026-09-23 18:04:39 +10:00
Luke Parker	fabf56781c	fix(app): keep timeline scroll when switching models (#50865 )	2026-09-23 18:03:42 +10:00
opencode	a25d304201	sync release versions for v2.0.15	2026-09-23 07:15:55 +00:00
Shoubhit Dash	cc8886c8bb	feat(ai): add video generation with queued media routes (#50703 )	2026-09-23 12:27:38 +05:30
Luke Parker	8ce629be22	fix(cli): keep Windows upgrades and uninstalls from fighting the running binary (#50819 )	2026-09-23 06:35:40 +00:00