@shop-os/modelkit 0.7.11 → 0.9.0

package/CHANGELOG.md

@@ -9,6 +9,33 @@ header in the release PR.
 
 ## [Unreleased]
 
+## [0.9.0] - Unreleased
+
+Fish Audio: text to speech, voice cloning and voice design (ENG-3400).
+
+### Added
+- Adapter `fish` for the Fish Audio REST API (`https://api.fish.audio`, base-url override `endpoints.fish`). Key `fish` from `FISH_AUDIO_API_KEY`, with the `ENTERPRISE_` / `SLOOSH_` prefixes; a sloosh call without `SLOOSH_FISH_AUDIO_API_KEY` fails with `auth`.
+- Speech rows `fish-s2.1-pro`, `fish-s2.1-pro-free` and `fish-drama-3` (the Fish `model` header `s2.1-pro`, `s2.1-pro-free`, `drama-3`). 24 Fish Official library voices (voice-designed or licensed, none a real-person clone; default Hannah), `mp3` (default) or `wav` output returned as bytes, a 10,000-character kit cap (Fish states none). Fish auto-detects the language, so the rows list none and `language` is dropped with a warning. Usage is one `images`-unit call, like the other speech rows; Fish bills per UTF-8 byte of the script ($15 / M bytes, Drama 3 $50 / M list, the free tier $0).
+- `SpeechRequest.customVoice`: a voice id `createVoice` returned, on rows with the new `capabilities.customVoices` (the Fish speech rows). It replaces `voice`; sending both, or a custom voice to another row, is `unsupported_option`.
+- Row types `voice-clone` and `voice-design`, and kit methods `createVoice` (row `fish-voice-clone`: 1-20 samples, optional transcripts and voice-design signatures on every sample or none, a private fast clone; returns `voiceId` and, when Fish reports a known one, `state`), `designVoice` (row `fish-voice-design-1`: a description, an optional preview line of up to 150 characters (the Fish OpenAPI limit), 1-4 candidates returned as WAV bytes with a `signature`; $0.01 per request) and `deleteVoice` (on the clone row; a 404 is `provider_rejected`, because Fish also answers 404 for a private voice of another account, except after a failed attempt of the same call, which may have deleted it). Usage unit `requests`. `resolveVoiceCloneOptions` / `resolveVoiceDesignOptions` and `resolveOptions` cover the new methods.
+- A cloned voice lives in the Fish account of the key set that made it: create, speak and delete with the same `tier`.
+- `constraints.voiceDesign` (`maxCandidates`, `previewMaxChars`).
+
+### Fixed
+- A 403 whose text says "Top up API credit" (Fish, out of API credit) is `provider_billing`, not `auth`.
+
+### Notes
+- Every Fish row ships with `exposure: []`: callable by id, listed nowhere, unpriced. Price and expose them in pro-backend per env when ENG-3400 is ready.
+- `createVoice` retries only a 429: after a 5xx Fish may already hold the voice, and a retry would make a second one.
+- A kit before 0.9.0 drops `voice-clone` / `voice-design` rows on load with a warning, and refuses the `fish` adapter id, so those rows stay invisible to consumers that have not upgraded.
+
+## [0.8.0] - Unreleased
+
+The BytePlus ModelArk key has one env name.
+
+### Removed
+- `keysFromEnv` reads the `ark` key from `BYTEDANCE_API_KEY` only (with the `ENTERPRISE_` / `SLOOSH_` prefixes); the two older alias names are no longer read. An env that sets only an alias gets no `ark` key, so an Ark call fails with a `ModelKitError` code `auth`. Set `BYTEDANCE_API_KEY` before a consumer upgrades.
+
 ## [0.7.11] - Unreleased
 
 Kling 4.0 on fal.

package/README.md

@@ -47,6 +47,12 @@ const synced = await kit.lipsync({ model: "lipsync-2", sourceVideo: { url }, aud
 const dubbed = await kit.dub({ model: "elevenlabs-dubbing", sourceVideo: { url }, targetLanguage: "es" });
 const transcript = await kit.transcribe({ model: "elevenlabs-scribe-v2", source: { url }, timestamps: true });
 
+// Custom voices (Fish Audio): clone once per workspace, keep `voiceId`, speak with it on the
+// same tier. `designVoice` returns candidates to keep the same way (pass `signature`).
+const { voiceId } = await kit.createVoice({ model: "fish-voice-clone", title, samples: [{ audio: { url } }] }, { tier });
+const mine = await kit.generateSpeech({ model: "fish-s2.1-pro", text, customVoice: voiceId }, { tier });
+await kit.deleteVoice({ model: "fish-voice-clone", voiceId }, { tier });
+
 const clip = await kit.prepareMediaPart(
     "gemini-2.5-flash",
     { bytes, mime: "video/mp4" },
@@ -108,7 +114,7 @@ before submit), and `model3dPricingSku(row, options, inputs)` names the flat pri
 bills under; `model3dSkus(row)` lists them all, and pro-backend keeps a `credit_pricing_flat`
 row for each. All three are in the browser-safe catalog entry.
 
-Keys come from `process.env` by default: bare names for pro (`OPENAI_API_KEY`, `FAL_KEY`, `ARK_API_KEY`, `LUMA_API_KEY` ...), the `ENTERPRISE_` prefix for enterprise and the `SLOOSH_` prefix for sloosh (sloosh reads `SLOOSH_LUMA_API_KEY` for Luma). Pass `keys: { pro, enterprise, sloosh }` to override. `tier: "sloosh"` uses the sloosh keys only: a missing key is a `ModelKitError` with code `auth`, never a fallback to pro.
+Keys come from `process.env` by default: bare names for pro (`OPENAI_API_KEY`, `FAL_KEY`, `BYTEDANCE_API_KEY`, `LUMA_API_KEY` ...), the `ENTERPRISE_` prefix for enterprise and the `SLOOSH_` prefix for sloosh (sloosh reads `SLOOSH_LUMA_API_KEY` for Luma). Pass `keys: { pro, enterprise, sloosh }` to override. `tier: "sloosh"` uses the sloosh keys only: a missing key is a `ModelKitError` with code `auth`, never a fallback to pro.
 
 ## Browser-safe catalog
 
@@ -344,7 +350,7 @@ the app's secrets, not a Cloudflare gateway token.
 
 ## Adapters
 
-`ai-sdk-text`, `google-image`, `openai-image`, `google-genai` (Veo, Omni with `delivery: "uri"`, Lyria 3, Files API media parts), `ark` (Seedream, Seedance), `fal`, `replicate` (images, video, Lyria 2, MMAudio), `prodia`, `higgsfield`, `xai-video`, `luma` (Uni images, Ray 3.2 video via the Agents API; `LUMA_API_KEY`, base-url override `endpoints.luma`). See `src/adapters/`.
+`ai-sdk-text`, `google-image`, `openai-image`, `google-genai` (Veo, Omni with `delivery: "uri"`, Lyria 3, Files API media parts), `ark` (Seedream, Seedance), `fal`, `replicate` (images, video, Lyria 2, MMAudio), `prodia`, `higgsfield`, `xai-video`, `luma` (Uni images, Ray 3.2 video via the Agents API; `LUMA_API_KEY`, base-url override `endpoints.luma`), `fish` (Fish Audio speech, `createVoice` / `designVoice` / `deleteVoice`; `FISH_AUDIO_API_KEY`, base-url override `endpoints.fish`). See `src/adapters/`.
 
 ## Not in the package
 

package/dist/catalog.cjs

@@ -74,6 +74,8 @@ __export(catalog_entry_exports, {
   resolveTranscriptionOptions: () => resolveTranscriptionOptions,
   resolveVideoOptions: () => resolveVideoOptions,
   resolveVoiceChangeOptions: () => resolveVoiceChangeOptions,
+  resolveVoiceCloneOptions: () => resolveVoiceCloneOptions,
+  resolveVoiceDesignOptions: () => resolveVoiceDesignOptions,
   sortCatalogRows: () => sortCatalogRows,
   takesTargetAspectRatio: () => takesTargetAspectRatio,
   videoLengthViolation: () => videoLengthViolation,
@@ -1302,6 +1304,303 @@ var VEED_SUBTITLES = {
   }
 };
 
+// src/catalog/voices-fish.ts
+var FISH_VOICES = [
+  {
+    id: "933563129e564b19a115bedd57b7406a",
+    name: "Sarah",
+    gender: "female",
+    languages: ["en"],
+    description: "An engaged speaker.",
+    previewUrl: "https://platform.r2.fish.audio/task/4e204945670c460c86fdab57512cb308.mp3"
+  },
+  {
+    id: "bf322df2096a46f18c579d0baa36f41d",
+    name: "Adrian",
+    gender: "male",
+    languages: ["en"],
+    description: "A steady and reliable narrator.",
+    previewUrl: "https://platform.r2.fish.audio/task/d730b09f4a484316a77ecc86137e85fb.mp3"
+  },
+  {
+    id: "536d3a5e000945adb7038665781a4aca",
+    name: "Ethan",
+    gender: "male",
+    languages: ["en"],
+    description: "A curious explainer.",
+    previewUrl: "https://platform.r2.fish.audio/task/bdd677ed767744fc91f166468786264b.mp3"
+  },
+  {
+    id: "b347db033a6549378b48d00acb0d06cd",
+    name: "Selene",
+    gender: "female",
+    languages: ["en"],
+    description: "A meditative female voice.",
+    previewUrl: "https://platform.r2.fish.audio/task/fcf629b4a0b343558f461338ba1b9da4.mp3"
+  },
+  {
+    id: "9a9cf47702da476aa4629e2506d4a857",
+    name: "Hannah",
+    gender: "female",
+    languages: ["en"],
+    description: "A conversation specialist.",
+    previewUrl: "https://platform.r2.fish.audio/task/08d658835feb427db808e37060b78eb0.mp3"
+  },
+  {
+    id: "79d0bd3e4e5444b18f7b6d89b5927bf1",
+    name: "Jordan",
+    gender: "male",
+    languages: ["en"],
+    description: "A motivational speaker.",
+    previewUrl: "https://platform.r2.fish.audio/task/edecbc0bdd4c46c39eda8520a0f729ef.mp3"
+  },
+  {
+    id: "e3cd384158934cc9a01029cd7d278634",
+    name: "Laura",
+    gender: "female",
+    languages: ["en"],
+    description: "A confident female narrator.",
+    previewUrl: "https://platform.r2.fish.audio/task/e821bef995344d519caa1c93d808db23.mp3"
+  },
+  {
+    id: "f3a2b90078d54a65af4b95f63bc7798e",
+    name: "Sheila",
+    gender: "female",
+    accent: "american",
+    languages: ["en"],
+    description: "A young American English woman's voice.",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-24/9bb9e1a1cf3f47f38700da1c5c0a678a.mp3"
+  },
+  {
+    id: "0db6e93af420470b9d9b9a886dbaa954",
+    name: "Kai",
+    gender: "male",
+    languages: ["en"],
+    description: "A friendly English man's voice for companion apps and everyday conversation.",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-24/13306f1172544c3ba3106d1e363feaa6.mp3"
+  },
+  {
+    id: "8abfa9ebd4c24a3399b68c3cdac8c7b5",
+    name: "Freya",
+    gender: "female",
+    accent: "british",
+    languages: ["en"],
+    description: "A natural, professional British English voice with a brisk, conversational pace. Clear and easy to follow, suited to voice assistants, customer support and everyday narration.",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-25/07918705a56b42daaab149f65392cea7.mp3"
+  },
+  {
+    id: "2f64e528a4f2406691925ab1e0ed5b7f",
+    name: "Ananya",
+    gender: "female",
+    languages: ["hi"],
+    description: "\u0938\u093E\u092B\u093C \u0914\u0930 \u0906\u0924\u094D\u092E\u0935\u093F\u0936\u094D\u0935\u093E\u0938 \u092D\u0930\u0940 \u0906\u0935\u093E\u091C\u093C, \u091C\u094B \u0935\u0949\u0907\u0938 \u0905\u0938\u093F\u0938\u094D\u091F\u0947\u0902\u091F, \u0917\u094D\u0930\u093E\u0939\u0915 \u0938\u0947\u0935\u093E \u0914\u0930 \u092A\u094D\u0930\u0938\u094D\u0924\u0941\u0924\u093F\u092F\u094B\u0902 \u0915\u0947 \u0932\u093F\u090F \u0909\u092A\u092F\u0941\u0915\u094D\u0924 \u0939\u0948\u0964",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-25/a17a700668d8405f8316db120733145c.mp3"
+  },
+  {
+    id: "cefd91a07b99468f9148e2580990082d",
+    name: "Arjun",
+    gender: "male",
+    languages: ["hi"],
+    description: "\u0938\u093E\u092B\u093C \u0914\u0930 \u0906\u0924\u094D\u092E\u0935\u093F\u0936\u094D\u0935\u093E\u0938 \u092D\u0930\u0940 \u0906\u0935\u093E\u091C\u093C, \u091C\u094B \u0935\u0949\u0907\u0938 \u0905\u0938\u093F\u0938\u094D\u091F\u0947\u0902\u091F, \u0917\u094D\u0930\u093E\u0939\u0915 \u0938\u0947\u0935\u093E \u0914\u0930 \u092A\u094D\u0930\u0938\u094D\u0924\u0941\u0924\u093F\u092F\u094B\u0902 \u0915\u0947 \u0932\u093F\u090F \u0909\u092A\u092F\u0941\u0915\u094D\u0924 \u0939\u0948\u0964",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-24/a0c4a8eadb7841cdae8e5e68031a2dfb.mp3"
+  },
+  {
+    id: "3f065d660212478a9c697fa965b4d098",
+    name: "Carmen",
+    gender: "female",
+    languages: ["es"],
+    description: "Una voz clara y segura para asistentes de voz, atenci\xF3n al cliente y presentaciones.",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-25/3421be48b6cb4c28bf829c25188e7e74.mp3"
+  },
+  {
+    id: "ec0718028a2445fb9be8e8bfb56a0150",
+    name: "Rodrigo",
+    gender: "male",
+    languages: ["es"],
+    description: "Una voz clara y segura para asistentes de voz, atenci\xF3n al cliente y presentaciones.",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-25/5a77c32365604800ad5b4eb13c5585df.mp3"
+  },
+  {
+    id: "10a3a20742114a4ea6dd441e7591850f",
+    name: "Manon",
+    gender: "female",
+    languages: ["fr"],
+    description: "Une voix claire et vive, adapt\xE9e aux assistants vocaux et au support client.",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-25/f7b75ee1fa524e48a8cefb003cf72952.mp3"
+  },
+  {
+    id: "f69bca092b674168a8d02d61ca20943c",
+    name: "Lucas",
+    gender: "male",
+    languages: ["fr"],
+    description: "Une voix claire et assur\xE9e, adapt\xE9e aux assistants vocaux et au service client.",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-25/86fca3e56a014a3d8a11450854bb8311.mp3"
+  },
+  {
+    id: "3937b44b604249f18b94913daef94043",
+    name: "Lena",
+    gender: "female",
+    languages: ["de"],
+    description: "Eine klare, souver\xE4ne Stimme f\xFCr Sprachassistenten, Kundenservice und Pr\xE4sentationen.",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-25/8659fec80b74442dab68a7755716a401.mp3"
+  },
+  {
+    id: "c1e7cc926aa5469aa966c353094d4b0b",
+    name: "Lukas",
+    gender: "male",
+    languages: ["de"],
+    description: "Eine klare, souver\xE4ne Stimme f\xFCr Sprachassistenten, Kundenservice und Pr\xE4sentationen.",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-25/fb1d538126bb4fe38f8366250f6a2136.mp3"
+  },
+  {
+    id: "b3c060cc41544a7da94a2c487647c902",
+    name: "Let\xEDcia",
+    gender: "female",
+    accent: "brazilian",
+    languages: ["pt"],
+    description: "A Brazilian-accented Portuguese woman's voice, in her late twenties. Soft, warm, intimate. A freelance designer from Belo Horizonte who enjoys long walks in the park. Best suited to companion apps and casual everyday conversation.",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-10-12/13c6c0d96b5c4069951f89868ee15360.mp3"
+  },
+  {
+    id: "add7bcb8d3c8423bbfa152ca232aad6e",
+    name: "Ricardo",
+    gender: "male",
+    accent: "brazilian",
+    languages: ["pt"],
+    description: "A Brazilian-accented Portuguese man's voice, in his late forties. Clear, professional, monotone. A high school history teacher from Curitiba who loves classical music. Best suited to lessons, tutoring and explainer content.",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-10-12/67475fc717ae45cbb38e647a2dd37fe1.mp3"
+  },
+  {
+    id: "b2d9d8db057042688a5e318b8f405bc2",
+    name: "Kyoko",
+    gender: "female",
+    languages: ["ja"],
+    description: "\u4E01\u5BE7\u3067\u843D\u3061\u7740\u3044\u305F\u8A71\u3057\u65B9\u304C\u7279\u5FB4\u306E\u5973\u6027\u306E\u58F0\u3002\u304A\u5BA2\u69D8\u5BFE\u5FDC\u3084\u60C5\u5831\u63D0\u4F9B\u306B\u6700\u9069\u3002",
+    previewUrl: "https://platform.r2.fish.audio/task/993b3ccb52124341aff79580f1ce5a5a.mp3"
+  },
+  {
+    id: "297a6fd278df47c3b9da9bfdf55ac89a",
+    name: "Satoru",
+    gender: "male",
+    languages: ["ja"],
+    description: "\u82E5\u3005\u3057\u304F\u3082\u843D\u3061\u7740\u304D\u306E\u3042\u308B\u7537\u6027\u306E\u58F0\u3002\u65E5\u5E38\u306E\u8A9E\u308A\u3084\u30EA\u30E9\u30C3\u30AF\u30B9\u3057\u305F\u30CA\u30EC\u30FC\u30B7\u30E7\u30F3\u306B\u6700\u9069\u3002",
+    previewUrl: "https://platform.r2.fish.audio/task/1b4712fb831f4c149eef6b0133157418.mp3"
+  },
+  {
+    id: "0b14f34a13a94ed88fe6113193749bb0",
+    name: "Fatima",
+    gender: "female",
+    languages: ["ar"],
+    description: "\u0635\u0648\u062A \u0648\u0627\u0636\u062D \u0648\u0648\u0627\u062B\u0642 \u064A\u0646\u0627\u0633\u0628 \u0627\u0644\u0645\u0633\u0627\u0639\u062F\u0627\u062A \u0627\u0644\u0635\u0648\u062A\u064A\u0629 \u0648\u062E\u062F\u0645\u0629 \u0627\u0644\u0639\u0645\u0644\u0627\u0621 \u0648\u0627\u0644\u0639\u0631\u0648\u0636 \u0627\u0644\u062A\u0642\u062F\u064A\u0645\u064A\u0629.",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-24/93406e1b905a4c4994306e3421243390.mp3"
+  },
+  {
+    id: "74c6aba5cbf94a15bbdc547ffce5cb38",
+    name: "Yutong",
+    gender: "female",
+    languages: ["zh"],
+    description: "\u58F0\u97F3\u6E05\u6670\u81EA\u7136\uFF0C\u8868\u8FBE\u5229\u843D\uFF0C\u9002\u5408\u8BED\u97F3\u52A9\u624B\u3001\u5BA2\u670D\u548C\u65E5\u5E38\u64AD\u62A5\u3002",
+    previewUrl: "https://platform.r2.fish.audio/tasks/2026-08-25/a48320f37fe44c4699549a1afc3778e8.mp3"
+  }
+];
+
+// src/catalog/speech-fish.ts
+var FISH_LOGO = "https://cdn.jsdelivr.net/npm/@lobehub/icons-static-png@1.97.0/light/fishaudio.png";
+var PER_CALL = { unit: "images" };
+var PER_REQUEST = { unit: "requests" };
+var TTS_MAX_CHARS = 1e4;
+function fishTts(id, label, providerModelId, description) {
+  return {
+    id,
+    aliases: [],
+    label,
+    type: "speech",
+    adapter: "fish",
+    providerModelId,
+    lifecycle: "active",
+    exposure: [],
+    constraints: {
+      voices: FISH_VOICES,
+      outputFormats: ["mp3", "wav"],
+      supports: [],
+      prompt: { referenceSyntax: "numbered", maxChars: TTS_MAX_CHARS }
+    },
+    defaults: { voice: "9a9cf47702da476aa4629e2506d4a857", outputFormat: "mp3" },
+    usage: PER_CALL,
+    pricingKind: "flat",
+    display: { provider: "fishaudio", logo: FISH_LOGO, description },
+    capabilities: { customVoices: true }
+  };
+}
+var FISH_MODELS = [
+  fishTts(
+    "fish-s2.1-pro",
+    "Fish Audio S2.1 Pro",
+    "s2.1-pro",
+    "Fish Audio S2.1 Pro text to speech in 83 languages, with [bracket] directions and your own cloned voices."
+  ),
+  fishTts(
+    "fish-s2.1-pro-free",
+    "Fish Audio S2.1 Pro Free",
+    "s2.1-pro-free",
+    "Fish Audio S2.1 Pro on the free developer tier: the same voices at no cost, without latency or data-processing guarantees."
+  ),
+  fishTts(
+    "fish-drama-3",
+    "Fish Audio Drama 3",
+    "drama-3",
+    "Fish Audio Drama 3 text to speech you direct like an actor: [directions], sounds, pauses and per-word tags."
+  ),
+  {
+    id: "fish-voice-clone",
+    aliases: [],
+    label: "Fish Audio Voice Clone",
+    type: "voice-clone",
+    adapter: "fish",
+    providerModelId: "voice-clone",
+    lifecycle: "active",
+    exposure: [],
+    constraints: {
+      // POST /model takes 1-20 voice files; Fish advises 10 s or more of clean speech each.
+      references: { audios: 20, audioFormats: ["wav", "mp3", "m4a", "opus"] },
+      supports: []
+    },
+    defaults: {},
+    usage: PER_REQUEST,
+    pricingKind: "flat",
+    display: {
+      provider: "fishaudio",
+      logo: FISH_LOGO,
+      description: "Clone a private, reusable voice from one or more clean recordings, then speak with it on the Fish Audio speech models."
+    },
+    capabilities: {}
+  },
+  {
+    id: "fish-voice-design-1",
+    aliases: [],
+    label: "Fish Audio Voice Design",
+    type: "voice-design",
+    adapter: "fish",
+    providerModelId: "voice-design-1",
+    lifecycle: "active",
+    exposure: [],
+    constraints: {
+      voiceDesign: { maxCandidates: 4, previewMaxChars: 150 },
+      supports: [],
+      prompt: { referenceSyntax: "numbered", maxChars: 2e3 }
+    },
+    defaults: {},
+    usage: PER_REQUEST,
+    pricingKind: "flat",
+    display: {
+      provider: "fishaudio",
+      logo: FISH_LOGO,
+      description: "Describe a voice and get up to four short candidates; keep one as a cloned voice."
+    },
+    capabilities: {}
+  }
+];
+
 // src/catalog/languages-audio.ts
 var ELEVENLABS_V4_LANGUAGES = [
   { code: "af", name: "Afrikaans" },
@@ -3383,7 +3682,7 @@ var ELEVENLABS_LOGO = "https://cdn.jsdelivr.net/npm/@lobehub/icons-static-png@1.
 var GOOGLE_LOGO = "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/192485d2-a76a-4657-9c06-930b2c70badd_compressed.png";
 var MINIMAX_LOGO = "https://d31vvfpk2ppdyq.cloudfront.net/assets/image/f4c2e906-28b7-4a4a-b491-df8fef7808ed.png";
 var SYNC_LOGO = "https://cdn.jsdelivr.net/npm/@lobehub/icons-static-png@1.97.0/light/sync.png";
-var PER_CALL = { unit: "images" };
+var PER_CALL2 = { unit: "images" };
 var EXPOSURE4 = ["cowork", "space_builder"];
 var LIPSYNC_AUDIO = { audioSeconds: { max: 30 } };
 function elevenLabsTts(tts) {
@@ -3404,7 +3703,7 @@ function elevenLabsTts(tts) {
       prompt: { referenceSyntax: "numbered", maxChars: tts.maxChars }
     },
     defaults: { voice: "Rachel", ...tts.formats && { outputFormat: "mp3_44100_128" } },
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: { provider: "elevenlabs", logo: ELEVENLABS_LOGO, description: tts.description },
     capabilities: {}
@@ -3428,7 +3727,7 @@ function geminiTts(id, label, description) {
       prompt: { referenceSyntax: "numbered", maxChars: 8192 }
     },
     defaults: { voice: "Kore" },
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: { provider: "google", logo: GOOGLE_LOGO, description },
     capabilities: {}
@@ -3451,7 +3750,7 @@ function syncLipsync(id, label, providerModelId, syncModel, description) {
     },
     defaults: {},
     ...syncModel !== void 0 && { providerOptions: { fal: { model: syncModel } } },
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: { provider: "sync", logo: SYNC_LOGO, description },
     capabilities: { acceptsSourceVideo: true }
@@ -3512,7 +3811,7 @@ var MEDIA_NODE_MODELS = [
       prompt: { referenceSyntax: "numbered", maxChars: 1e4 }
     },
     defaults: { voice: "Wise_Woman" },
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: {
       provider: "minimax",
@@ -3539,7 +3838,7 @@ var MEDIA_NODE_MODELS = [
       supports: []
     },
     defaults: { voice: "Rachel", outputFormat: "mp3_44100_128" },
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: {
       provider: "elevenlabs",
@@ -3566,7 +3865,7 @@ var MEDIA_NODE_MODELS = [
       supports: []
     },
     defaults: {},
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: {
       provider: "elevenlabs",
@@ -3613,7 +3912,7 @@ var MEDIA_NODE_MODELS = [
       supports: []
     },
     defaults: {},
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: {
       provider: "veed",
@@ -3665,7 +3964,7 @@ var MEDIA_NODE_MODELS = [
       supports: []
     },
     defaults: {},
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: {
       provider: "elevenlabs",
@@ -3674,7 +3973,8 @@ var MEDIA_NODE_MODELS = [
     },
     capabilities: {}
   },
-  SAM_3_IMAGE
+  SAM_3_IMAGE,
+  ...FISH_MODELS
 ];
 
 // src/catalog/model3d.ts
@@ -7315,7 +7615,8 @@ var ADAPTER_IDS = [
   "prodia",
   "higgsfield",
   "xai-video",
-  "luma"
+  "luma",
+  "fish"
 ];
 var ModelTypeSchema = import_zod3.z.enum([
   "text",
@@ -7332,6 +7633,8 @@ var ModelTypeSchema = import_zod3.z.enum([
   "dubbing",
   "lipsync",
   "transcription",
+  "voice-clone",
+  "voice-design",
   "model3d"
 ]);
 var LifecycleSchema = import_zod3.z.enum(["active", "deprecated", "retired"]);
@@ -7507,6 +7810,10 @@ var ModelConstraintsSchema = import_zod3.z.looseObject({
   supports: import_zod3.z.array(SupportedOptionSchema).optional(),
   ratioIgnoredWith: import_zod3.z.array(import_zod3.z.enum(["firstFrame", "sourceVideo"])).optional(),
   voices: import_zod3.z.array(VoiceOptionSchema).optional(),
+  voiceDesign: import_zod3.z.object({
+    maxCandidates: import_zod3.z.number().int().positive(),
+    previewMaxChars: import_zod3.z.number().int().positive()
+  }).optional(),
   languages: import_zod3.z.array(LanguageOptionSchema).optional(),
   outputFormats: import_zod3.z.array(import_zod3.z.string().min(1)).min(1).optional(),
   sourceVideoSeconds: VideoSecondsLimitsSchema.omit({ total: true }).extend({ inReferenceTotal: import_zod3.z.boolean().optional() }).optional(),
@@ -7605,6 +7912,7 @@ var CatalogDisplaySchema = import_zod3.z.looseObject({
 });
 var CatalogCapabilitiesSchema = import_zod3.z.looseObject({
   acceptsSourceVideo: import_zod3.z.boolean().optional(),
+  customVoices: import_zod3.z.boolean().optional(),
   videoEdit: import_zod3.z.boolean().optional(),
   framesExcludeReferences: import_zod3.z.boolean().optional(),
   sourceVideoExcludesFrames: import_zod3.z.boolean().optional(),
@@ -9483,8 +9791,17 @@ function resolveSpeechOptions(model, request, onUnsupported = "error") {
   const warnings = [];
   checkScript(model, request.text);
   const resolved = { model: model.id, text: request.text };
-  const voice = resolveVoice(model, request.voice, onUnsupported, warnings);
-  if (voice !== void 0) resolved.voice = voice;
+  const customVoice = request.customVoice?.trim();
+  if (customVoice) {
+    if (model.capabilities?.customVoices !== true) {
+      throw unsupported3(model, "does not take a custom voice");
+    }
+    if (request.voice?.trim()) throw unsupported3(model, "takes voice or customVoice, not both");
+    resolved.customVoice = customVoice;
+  } else {
+    const voice = resolveVoice(model, request.voice, onUnsupported, warnings);
+    if (voice !== void 0) resolved.voice = voice;
+  }
   const language = resolveListed(
     model,
     {
@@ -9606,6 +9923,106 @@ function resolveTranscriptionOptions(model, request, onUnsupported = "error") {
   return { request: resolved, warnings };
 }
 
+// src/resolve-voice.ts
+function refuse(model, code, message) {
+  return new ModelKitError({ code, message: `${model.label} ${message}`, model: model.id });
+}
+function checkSample(model, sample, index) {
+  const limits = model.constraints.references?.audioSeconds;
+  const length = sample.audio.durationSeconds;
+  if (limits === void 0 || length === void 0) return;
+  if (limits.min !== void 0 && length < limits.min || limits.max !== void 0 && length > limits.max) {
+    const range2 = `${limits.min ?? 0}-${limits.max ?? "any"} s`;
+    throw refuse(
+      model,
+      "unsupported_option",
+      `takes samples of ${range2}; sample ${index + 1} is ${length} s`
+    );
+  }
+}
+function resolveVoiceCloneOptions(model, request) {
+  const title = request.title.trim();
+  if (!title) throw refuse(model, "missing_input", "needs a title");
+  if (request.samples.length === 0) throw refuse(model, "missing_input", "needs a voice sample");
+  const max = model.constraints.references?.audios;
+  if (max !== void 0 && request.samples.length > max) {
+    throw refuse(
+      model,
+      "unsupported_option",
+      `takes up to ${max} samples; got ${request.samples.length}`
+    );
+  }
+  request.samples.forEach((sample, index) => checkSample(model, sample, index));
+  const total = model.constraints.references?.audioSeconds?.total;
+  const seconds3 = request.samples.reduce((sum, s) => sum + (s.audio.durationSeconds ?? 0), 0);
+  if (total !== void 0 && seconds3 > total) {
+    throw refuse(
+      model,
+      "unsupported_option",
+      `takes up to ${total} s of samples; got ${seconds3} s`
+    );
+  }
+  for (const field of ["transcript", "designSignature"]) {
+    const set = request.samples.filter((s) => s[field]?.trim()).length;
+    if (set !== 0 && set !== request.samples.length) {
+      throw refuse(model, "missing_input", `needs ${field} on every sample or on none`);
+    }
+  }
+  return { request: { ...request, model: model.id, title }, warnings: [] };
+}
+function resolveVoiceDesignOptions(model, request, onUnsupported = "error") {
+  const warnings = [];
+  const prompt = request.prompt.trim();
+  if (!prompt) throw refuse(model, "missing_input", "needs a voice description");
+  const maxChars = promptConventions(model).maxChars;
+  if (maxChars !== void 0 && prompt.length > maxChars) {
+    throw refuse(
+      model,
+      "unsupported_option",
+      `takes up to ${maxChars} characters; got ${prompt.length}`
+    );
+  }
+  const limits = model.constraints.voiceDesign;
+  const resolved = { model: model.id, prompt };
+  const preview2 = request.previewText?.trim();
+  if (preview2) {
+    if (limits !== void 0 && preview2.length > limits.previewMaxChars) {
+      throw refuse(
+        model,
+        "unsupported_option",
+        `takes a preview line of up to ${limits.previewMaxChars} characters; got ${preview2.length}`
+      );
+    }
+    resolved.previewText = preview2;
+  }
+  const language = request.language?.trim();
+  if (language) resolved.language = language;
+  if (request.seed !== void 0) resolved.seed = request.seed;
+  const count2 = request.count ?? model.defaults.count;
+  if (count2 !== void 0) {
+    const cap = limits?.maxCandidates ?? count2;
+    if (!Number.isInteger(count2) || count2 < 1 || count2 > cap) {
+      if (onUnsupported === "error" || !Number.isInteger(count2)) {
+        throw refuse(
+          model,
+          "unsupported_option",
+          `returns 1-${cap} candidates; got ${count2}`
+        );
+      }
+      const to = Math.min(Math.max(count2, 1), cap);
+      warnings.push({
+        code: "option_ignored",
+        field: "count",
+        from: String(count2),
+        to: String(to),
+        message: `count ${count2} replaced by ${to} for ${model.id}`
+      });
+      resolved.count = to;
+    } else resolved.count = count2;
+  }
+  return { request: resolved, warnings };
+}
+
 // src/resolve-call.ts
 var MEDIA_METHODS = {
   text: [],
@@ -9621,6 +10038,9 @@ var MEDIA_METHODS = {
   lipsync: ["lipsync"],
   dubbing: ["dub"],
   transcription: ["transcribe"],
+  // deleteVoice runs on the voice-clone row too; it checks with assertMethod("createVoice").
+  "voice-clone": ["createVoice"],
+  "voice-design": ["designVoice"],
   "video-tool": ["submitVideoTool"],
   // 3D jobs run through submitModel3d, which checks the row type itself; their preflight
   // is model3dInputRefusal, not resolveOptions.
@@ -9635,6 +10055,8 @@ var METHOD_NOUN = {
   lipsync: "a lipsync model",
   dub: "a dubbing model",
   transcribe: "a transcription model",
+  createVoice: "a voice cloning model",
+  designVoice: "a voice design model",
   submitVideoTool: "a video tool"
 };
 function assertMethod(model, method) {
@@ -9677,6 +10099,14 @@ function resolveTranscriptionCall(model, request, onUnsupported = "error") {
   assertMethod(model, "transcribe");
   return resolveTranscriptionOptions(model, request, onUnsupported);
 }
+function resolveVoiceCloneCall(model, request) {
+  assertMethod(model, "createVoice");
+  return resolveVoiceCloneOptions(model, request);
+}
+function resolveVoiceDesignCall(model, request, onUnsupported = "error") {
+  assertMethod(model, "designVoice");
+  return resolveVoiceDesignOptions(model, request, onUnsupported);
+}
 function resolveVideoToolCall(model, request) {
   assertMethod(model, "submitVideoTool");
   const plan = planVideoTool(model, request);
@@ -9723,6 +10153,13 @@ function resolveOptions(model, call, onUnsupported = "error") {
         method: call.method,
         ...resolveTranscriptionCall(model, call.request, onUnsupported)
       };
+    case "createVoice":
+      return { method: call.method, ...resolveVoiceCloneCall(model, call.request) };
+    case "designVoice":
+      return {
+        method: call.method,
+        ...resolveVoiceDesignCall(model, call.request, onUnsupported)
+      };
     case "submitVideoTool":
       return { method: call.method, ...resolveVideoToolCall(model, call.request) };
   }