@shop-os/modelkit 0.8.0 → 0.9.0

package/dist/index.js

@@ -1230,6 +1230,303 @@ 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" },
@@ -3311,7 +3608,7 @@ 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) {
@@ -3332,7 +3629,7 @@ 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: {}
@@ -3356,7 +3653,7 @@ 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: {}
@@ -3379,7 +3676,7 @@ 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 }
@@ -3440,7 +3737,7 @@ var MEDIA_NODE_MODELS = [
       prompt: { referenceSyntax: "numbered", maxChars: 1e4 }
     },
     defaults: { voice: "Wise_Woman" },
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: {
       provider: "minimax",
@@ -3467,7 +3764,7 @@ var MEDIA_NODE_MODELS = [
       supports: []
     },
     defaults: { voice: "Rachel", outputFormat: "mp3_44100_128" },
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: {
       provider: "elevenlabs",
@@ -3494,7 +3791,7 @@ var MEDIA_NODE_MODELS = [
       supports: []
     },
     defaults: {},
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: {
       provider: "elevenlabs",
@@ -3541,7 +3838,7 @@ var MEDIA_NODE_MODELS = [
       supports: []
     },
     defaults: {},
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: {
       provider: "veed",
@@ -3593,7 +3890,7 @@ var MEDIA_NODE_MODELS = [
       supports: []
     },
     defaults: {},
-    usage: PER_CALL,
+    usage: PER_CALL2,
     pricingKind: "flat",
     display: {
       provider: "elevenlabs",
@@ -3602,7 +3899,8 @@ var MEDIA_NODE_MODELS = [
     },
     capabilities: {}
   },
-  SAM_3_IMAGE
+  SAM_3_IMAGE,
+  ...FISH_MODELS
 ];
 
 // src/catalog/model3d.ts
@@ -7237,7 +7535,8 @@ var ADAPTER_IDS = [
   "prodia",
   "higgsfield",
   "xai-video",
-  "luma"
+  "luma",
+  "fish"
 ];
 var ModelTypeSchema = z3.enum([
   "text",
@@ -7254,6 +7553,8 @@ var ModelTypeSchema = z3.enum([
   "dubbing",
   "lipsync",
   "transcription",
+  "voice-clone",
+  "voice-design",
   "model3d"
 ]);
 var LifecycleSchema = z3.enum(["active", "deprecated", "retired"]);
@@ -7429,6 +7730,10 @@ var ModelConstraintsSchema = z3.looseObject({
   supports: z3.array(SupportedOptionSchema).optional(),
   ratioIgnoredWith: z3.array(z3.enum(["firstFrame", "sourceVideo"])).optional(),
   voices: z3.array(VoiceOptionSchema).optional(),
+  voiceDesign: z3.object({
+    maxCandidates: z3.number().int().positive(),
+    previewMaxChars: z3.number().int().positive()
+  }).optional(),
   languages: z3.array(LanguageOptionSchema).optional(),
   outputFormats: z3.array(z3.string().min(1)).min(1).optional(),
   sourceVideoSeconds: VideoSecondsLimitsSchema.omit({ total: true }).extend({ inReferenceTotal: z3.boolean().optional() }).optional(),
@@ -7527,6 +7832,7 @@ var CatalogDisplaySchema = z3.looseObject({
 });
 var CatalogCapabilitiesSchema = z3.looseObject({
   acceptsSourceVideo: z3.boolean().optional(),
+  customVoices: z3.boolean().optional(),
   videoEdit: z3.boolean().optional(),
   framesExcludeReferences: z3.boolean().optional(),
   sourceVideoExcludesFrames: z3.boolean().optional(),
@@ -8046,7 +8352,7 @@ var errorCodes = z6.object({
   type: z6.string().optional(),
   error: z6.object({ code: z6.string().optional(), type: z6.string().optional() }).optional()
 });
-var BILLING_TEXT = /insufficient_quota|payment[ _]required|exhausted balance|insufficient (?:credits?|balance|funds)|billing_hard_limit/i;
+var BILLING_TEXT = /insufficient_quota|payment[ _]required|exhausted balance|insufficient (?:credits?|balance|funds)|billing_hard_limit|top up (?:api )?credit/i;
 var BILLING_TEXT_STATUSES = /* @__PURE__ */ new Set([403, 429]);
 function billingCode(body) {
   const parsed = errorCodes.safeParse(body);
@@ -8443,16 +8749,16 @@ async function generateSeedreamImage(req, ctx) {
   } else {
     throwArkError(void 0, "Seedream returned no image URL or base64 data", modelId);
   }
-  const usage3 = {
+  const usage4 = {
     model: modelId,
     provider: "ark",
     modality: "IMAGE",
     unit: "images",
     quantity: 1
   };
-  if (req.size !== void 0) usage3.imageSize = req.size;
-  if (req.quality !== void 0) usage3.imageQuality = req.quality;
-  return { images: [{ bytes, mime: sniffedImageMime(bytes) }], usage: usage3, raw: data };
+  if (req.size !== void 0) usage4.imageSize = req.size;
+  if (req.quality !== void 0) usage4.imageQuality = req.quality;
+  return { images: [{ bytes, mime: sniffedImageMime(bytes) }], usage: usage4, raw: data };
 }
 var SeedanceUsageSchema = z7.object({
   total_tokens: z7.number().optional(),
@@ -8493,7 +8799,7 @@ function seedanceErrorMessage(error, status) {
   if (asObject.success && asObject.data.message !== void 0) return asObject.data.message;
   return `Seedance task ${status}`;
 }
-function seedanceJob(ctx, key, jobId, usage3) {
+function seedanceJob(ctx, key, jobId, usage4) {
   const modelId = ctx.model.id;
   const fetchStatus = async () => {
     const res = await arkFetch(ctx, `${arkBaseUrl(ctx)}${SUBMIT_PATH}/${jobId}`, {
@@ -8543,21 +8849,21 @@ function seedanceJob(ctx, key, jobId, usage3) {
               modelId
             );
           if (polled.duration !== void 0 && polled.duration > 0) {
-            usage3.quantity = polled.duration;
-            usage3.durationSeconds = polled.duration;
+            usage4.quantity = polled.duration;
+            usage4.durationSeconds = polled.duration;
           }
           const rendered = VideoResolutionSchema.safeParse(
             polled.resolution?.toLowerCase()
           );
-          if (rendered.success) usage3.resolution = rendered.data;
+          if (rendered.success) usage4.resolution = rendered.data;
           const tokens = polled.usage;
           if (tokens?.total_tokens !== void 0 && tokens.completion_tokens !== void 0) {
-            usage3.inputTokens = tokens.total_tokens - tokens.completion_tokens;
-            usage3.outputTokens = tokens.completion_tokens;
+            usage4.inputTokens = tokens.total_tokens - tokens.completion_tokens;
+            usage4.outputTokens = tokens.completion_tokens;
           }
           return {
             video: { url, expiresAt: new Date(Date.now() + VIDEO_URL_TTL_MS) },
-            usage: usage3,
+            usage: usage4,
             raw: polled
           };
         }
@@ -8663,8 +8969,8 @@ async function generateSeedanceVideo(req, ctx) {
   );
   if (!submit.id) throwArkError(void 0, "Seedance task create returned no id", modelId);
   const jobId = submit.id;
-  const usage3 = seedanceUsage(ctx, req, taskType === "edit" ? requestedDuration : duration);
-  return seedanceJob(ctx, key, jobId, usage3);
+  const usage4 = seedanceUsage(ctx, req, taskType === "edit" ? requestedDuration : duration);
+  return seedanceJob(ctx, key, jobId, usage4);
 }
 
 // src/adapters/fal.ts
@@ -10152,7 +10458,7 @@ async function collectFalPlanImages(payload, plan4, ctx, req) {
       return { bytes, mime: mime || "image/png" };
     })
   );
-  const usage3 = {
+  const usage4 = {
     model: ctx.model.id,
     provider: "fal",
     modality: "IMAGE",
@@ -10160,9 +10466,9 @@ async function collectFalPlanImages(payload, plan4, ctx, req) {
     quantity: images.length
   };
   const billedSize = plan4.imageSize ?? req.size;
-  if (billedSize !== void 0 && ctx.model.type === "image") usage3.imageSize = billedSize;
-  if (plan4.imageQuality !== void 0) usage3.imageQuality = plan4.imageQuality;
-  const result = { images, usage: usage3, raw: payload };
+  if (billedSize !== void 0 && ctx.model.type === "image") usage4.imageSize = billedSize;
+  if (plan4.imageQuality !== void 0) usage4.imageQuality = plan4.imageQuality;
+  const result = { images, usage: usage4, raw: payload };
   const remote = files.find((f) => !f.url.startsWith("data:"));
   if (remote !== void 0) result.outputUrl = remote.url;
   return result;
@@ -12325,14 +12631,297 @@ function planUsage(ctx, plan4) {
   };
 }
 
+// src/adapters/fish.ts
+import { z as z21 } from "zod";
+var DEFAULT_BASE_URL = "https://api.fish.audio";
+var FISH_TTS_MODEL_IDS = /* @__PURE__ */ new Set([
+  "s2.1-pro",
+  "s2.1-pro-free",
+  "drama-3"
+]);
+var FISH_VOICE_CLONE = "voice-clone";
+var FISH_VOICE_DESIGN = "voice-design-1";
+var VOICE_ID = /^[A-Za-z0-9_-]{1,64}$/;
+function fishRowRefusal(row2) {
+  const id = row2.providerModelId;
+  if (row2.type === "speech") {
+    return FISH_TTS_MODEL_IDS.has(id) ? void 0 : `fish speech rows need providerModelId ${[...FISH_TTS_MODEL_IDS].join(", ")}`;
+  }
+  if (row2.type === "voice-clone") {
+    return id === FISH_VOICE_CLONE ? void 0 : `fish voice-clone rows need providerModelId ${FISH_VOICE_CLONE}`;
+  }
+  if (row2.type === "voice-design") {
+    return id === FISH_VOICE_DESIGN ? void 0 : `fish voice-design rows need providerModelId ${FISH_VOICE_DESIGN}`;
+  }
+  return `fish has no ${row2.type} rows`;
+}
+function fishError(ctx, code, message) {
+  return new ModelKitError({
+    code,
+    message,
+    provider: "fish",
+    model: ctx.model.id,
+    retryable: false
+  });
+}
+function checkRow(ctx, type) {
+  const refusal2 = ctx.model.type === type ? fishRowRefusal(ctx.model) : void 0;
+  if (ctx.model.type !== type || refusal2 !== void 0) {
+    throw fishError(
+      ctx,
+      "unsupported_option",
+      refusal2 ?? `${ctx.model.id} is not a ${type} row`
+    );
+  }
+}
+function checkVoiceId(ctx, field, id) {
+  if (!VOICE_ID.test(id))
+    throw fishError(ctx, "unsupported_option", `${field} is not a voice id`);
+  return id;
+}
+function fishKey(ctx) {
+  const key = ctx.keys.fish;
+  if (!key) throw fishError(ctx, "auth", "Missing API key for fish");
+  return key;
+}
+function fishUrl(ctx, path) {
+  return `${ctx.endpoints?.fish ?? DEFAULT_BASE_URL}${path}`;
+}
+async function fishFetch(ctx, path, init) {
+  throwIfAborted(ctx.signal);
+  const request = {
+    method: init.method,
+    headers: { Authorization: `Bearer ${fishKey(ctx)}`, ...init.headers }
+  };
+  if (init.body !== void 0) request.body = init.body;
+  if (ctx.signal !== void 0) request.signal = ctx.signal;
+  let res;
+  try {
+    res = await ctx.fetch(fishUrl(ctx, path), request);
+  } catch (cause) {
+    throw classifyThrown("fish", cause, ctx.model.id);
+  }
+  if (!res.ok) throw classifyHttpError(res.status, await readBody(res), "fish", ctx.model.id);
+  return res;
+}
+function usage2(ctx) {
+  return {
+    model: ctx.model.id,
+    provider: "fish",
+    modality: "AUDIO",
+    unit: ctx.model.usage.unit,
+    quantity: 1
+  };
+}
+function audioMime(res, format) {
+  const header = res.headers.get("content-type")?.split(";")[0]?.trim().toLowerCase();
+  if (header?.startsWith("audio/")) return header;
+  return format === "wav" ? "audio/wav" : "audio/mpeg";
+}
+async function generateSpeech(req, ctx) {
+  checkRow(ctx, "speech");
+  const format = req.outputFormat ?? "mp3";
+  const body = { text: req.text, format };
+  const voice = req.customVoice ?? req.voice;
+  if (voice !== void 0) {
+    body.reference_id = checkVoiceId(ctx, req.customVoice ? "customVoice" : "voice", voice);
+  }
+  if (req.language !== void 0) body.language = req.language;
+  const { bytes, mime } = await withRetry(
+    async () => {
+      const res = await fishFetch(ctx, "/v1/tts", {
+        method: "POST",
+        headers: { "Content-Type": "application/json", model: ctx.model.providerModelId },
+        body: JSON.stringify(body)
+      });
+      let audio;
+      try {
+        audio = new Uint8Array(await res.arrayBuffer());
+      } catch (cause) {
+        throw classifyThrown("fish", cause, ctx.model.id);
+      }
+      return { bytes: audio, mime: audioMime(res, format) };
+    },
+    mediaRetryOptions(ctx.call, ctx.signal)
+  );
+  if (bytes.byteLength === 0)
+    throw fishError(ctx, "provider_rejected", "Fish TTS returned no audio");
+  return { audio: { bytes, mime }, usage: usage2(ctx) };
+}
+var createdModelSchema = z21.object({
+  _id: z21.string().min(1),
+  state: z21.enum(["created", "training", "trained", "failed"]).optional().catch(void 0)
+});
+function fileName(index, mime) {
+  const ext = mime.split("/")[1]?.replace("mpeg", "mp3").replace("x-", "") || "bin";
+  return `sample-${index + 1}.${ext}`;
+}
+function allOrNone(ctx, field, values) {
+  const set = values.filter((v) => v !== void 0 && v.trim() !== "");
+  if (set.length === 0) return [];
+  if (set.length !== values.length) {
+    throw fishError(ctx, "missing_input", `set ${field} on every sample or on none`);
+  }
+  return set;
+}
+async function cloneForm(req, ctx) {
+  if (req.samples.length === 0) throw fishError(ctx, "missing_input", "needs a voice sample");
+  const transcripts = allOrNone(
+    ctx,
+    "transcript",
+    req.samples.map((s) => s.transcript)
+  );
+  const signatures = allOrNone(
+    ctx,
+    "designSignature",
+    req.samples.map((s) => s.designSignature)
+  );
+  const form = new FormData();
+  form.append("type", "tts");
+  form.append("train_mode", "fast");
+  form.append("visibility", "private");
+  form.append("title", req.title.trim());
+  if (req.description?.trim()) form.append("description", req.description.trim());
+  if (req.enhance !== void 0) form.append("enhance_audio_quality", String(req.enhance));
+  for (const [index, sample] of req.samples.entries()) {
+    const { bytes, mime } = await mediaRefToBytes(sample.audio, ctx.fetch, {
+      signal: ctx.signal,
+      provider: "fish",
+      model: ctx.model.id
+    });
+    const audio = new Blob([new Uint8Array(bytes)], { type: mime });
+    form.append("voices", audio, fileName(index, mime));
+  }
+  for (const text2 of transcripts) form.append("texts", text2.trim());
+  for (const signature of signatures) form.append("voice_design_signatures", signature);
+  return form;
+}
+async function onlyRateLimitRetries(fn) {
+  try {
+    return await fn();
+  } catch (cause) {
+    if (!isModelKitError(cause) || !cause.retryable || cause.code === "rate_limited")
+      throw cause;
+    throw new ModelKitError({
+      code: cause.code,
+      message: cause.message,
+      ...cause.provider !== void 0 && { provider: cause.provider },
+      ...cause.model !== void 0 && { model: cause.model },
+      ...cause.status !== void 0 && { status: cause.status },
+      ...cause.body !== void 0 && { body: cause.body },
+      cause,
+      retryable: false
+    });
+  }
+}
+async function createVoice(req, ctx) {
+  checkRow(ctx, "voice-clone");
+  const form = await cloneForm(req, ctx);
+  const res = await withRetry(
+    () => onlyRateLimitRetries(() => fishFetch(ctx, "/model", { method: "POST", body: form })),
+    mediaRetryOptions(ctx.call, ctx.signal)
+  );
+  const raw = await readBody(res);
+  const parsed = createdModelSchema.safeParse(raw);
+  if (!parsed.success) {
+    throw new ModelKitError({
+      code: "provider_rejected",
+      message: "Fish returned no voice id",
+      provider: "fish",
+      model: ctx.model.id,
+      status: res.status,
+      body: raw,
+      retryable: false
+    });
+  }
+  const result = { voiceId: parsed.data._id, usage: usage2(ctx), raw };
+  if (parsed.data.state !== void 0) result.state = parsed.data.state;
+  return result;
+}
+var designSchema = z21.object({
+  candidates: z21.array(
+    z21.looseObject({
+      id: z21.string().min(1),
+      index: z21.number().int(),
+      audio_base64: z21.string().min(1),
+      duration_ms: z21.number().nonnegative(),
+      signature: z21.string().min(1).optional()
+    })
+  ).min(1)
+});
+async function designVoice(req, ctx) {
+  checkRow(ctx, "voice-design");
+  const body = { instruction: req.prompt.trim() };
+  if (req.previewText?.trim()) body.reference_text = req.previewText.trim();
+  if (req.language !== void 0) body.language = req.language;
+  if (req.count !== void 0) body.n = req.count;
+  if (req.seed !== void 0) body.seed = req.seed;
+  const raw = await withRetry(
+    async () => {
+      const res = await fishFetch(ctx, "/v1/voice-design", {
+        method: "POST",
+        headers: { "Content-Type": "application/json", model: ctx.model.providerModelId },
+        body: JSON.stringify(body)
+      });
+      return readBody(res);
+    },
+    mediaRetryOptions(ctx.call, ctx.signal)
+  );
+  const parsed = designSchema.safeParse(raw);
+  if (!parsed.success) {
+    throw fishError(ctx, "provider_rejected", "Fish voice design returned no candidates");
+  }
+  const candidates = [...parsed.data.candidates].sort((a, b) => a.index - b.index).map((c) => ({
+    audio: { bytes: base64ToBytes(c.audio_base64), mime: "audio/wav" },
+    durationSeconds: c.duration_ms / 1e3,
+    // The OpenAPI candidate has no signature field, so its stable id is the handle; Fish
+    // rejects a clone whose signature does not verify, so a wrong one fails loudly.
+    signature: c.signature ?? c.id
+  }));
+  return { candidates, usage: usage2(ctx) };
+}
+async function deleteVoice(req, ctx) {
+  checkRow(ctx, "voice-clone");
+  const id = encodeURIComponent(checkVoiceId(ctx, "voiceId", req.voiceId));
+  let attempts = 0;
+  await withRetry(
+    async () => {
+      attempts += 1;
+      try {
+        await fishFetch(ctx, `/model/${id}`, { method: "DELETE" });
+      } catch (cause) {
+        if (!isModelKitError(cause) || cause.status !== 404) throw cause;
+        if (attempts > 1) return;
+        throw new ModelKitError({
+          code: "provider_rejected",
+          message: `Fish has no voice ${req.voiceId} under this key set; delete with the tier that created it`,
+          provider: "fish",
+          model: ctx.model.id,
+          status: 404,
+          cause,
+          retryable: false
+        });
+      }
+    },
+    mediaRetryOptions(ctx.call, ctx.signal)
+  );
+}
+var fishAdapter = {
+  id: "fish",
+  generateSpeech,
+  createVoice,
+  designVoice,
+  deleteVoice
+};
+
 // src/adapters/google-genai.ts
-import { z as z22 } from "zod";
+import { z as z23 } from "zod";
 import {
   GoogleGenAI
 } from "@google/genai";
 
 // src/adapters/google-files.ts
-import { z as z21 } from "zod";
+import { z as z22 } from "zod";
 var GOOGLE_FILES_BASE = "https://generativelanguage.googleapis.com";
 var GOOGLE_FILES_POLL_INTERVAL_MS = 2e3;
 var GOOGLE_FILES_POLL_TIMEOUT_MS = 5 * 60 * 1e3;
@@ -12342,17 +12931,17 @@ var GENERIC_MEDIA_TYPES = /* @__PURE__ */ new Set([
   "binary/octet-stream",
   "application/binary"
 ]);
-var uploadedFileSchema = z21.object({
-  file: z21.object({
-    uri: z21.string().optional(),
-    name: z21.string().optional(),
-    state: z21.string().optional(),
-    mimeType: z21.string().optional()
+var uploadedFileSchema = z22.object({
+  file: z22.object({
+    uri: z22.string().optional(),
+    name: z22.string().optional(),
+    state: z22.string().optional(),
+    mimeType: z22.string().optional()
   }).optional()
 });
-var fileStateSchema = z21.object({
-  state: z21.string().optional(),
-  error: z21.object({ message: z21.string().optional() }).optional()
+var fileStateSchema = z22.object({
+  state: z22.string().optional(),
+  error: z22.object({ message: z22.string().optional() }).optional()
 });
 function resolveMime(declared, fallback) {
   const clean = declared?.split(";")[0]?.trim().toLowerCase();
@@ -12459,7 +13048,7 @@ async function startAndUpload(bytes, mimeType, opts) {
   if (uploaded.file?.state !== void 0) ref.state = uploaded.file.state;
   return ref;
 }
-async function waitForActive(fileName, opts) {
+async function waitForActive(fileName2, opts) {
   const pollOptions2 = {
     intervalMs: GOOGLE_FILES_POLL_INTERVAL_MS,
     timeoutMs: GOOGLE_FILES_POLL_TIMEOUT_MS
@@ -12472,14 +13061,14 @@ async function waitForActive(fileName, opts) {
     retryInit.signal = opts.signal;
   }
   await pollUntil(async () => {
-    const res = await opts.fetchImpl(`${GOOGLE_FILES_BASE}/v1beta/${fileName}`, retryInit);
+    const res = await opts.fetchImpl(`${GOOGLE_FILES_BASE}/v1beta/${fileName2}`, retryInit);
     if (res.ok) {
       const json = fileStateSchema.parse(await res.json());
       if (json.state === "ACTIVE") return true;
       if (json.state === "FAILED") {
         throw classifyHttpError(
           void 0,
-          `Files API processing failed for ${fileName}: ${json.error?.message ?? "unknown error"}`,
+          `Files API processing failed for ${fileName2}: ${json.error?.message ?? "unknown error"}`,
           "google-genai",
           opts.model
         );
@@ -12491,9 +13080,9 @@ async function waitForActive(fileName, opts) {
     throw classifyHttpError(res.status, text2, "google-genai", opts.model);
   }, pollOptions2);
 }
-async function deleteGoogleFile(fileName, apiKey, fetchImpl) {
+async function deleteGoogleFile(fileName2, apiKey, fetchImpl) {
   try {
-    await fetchImpl(`${GOOGLE_FILES_BASE}/v1beta/${fileName}`, {
+    await fetchImpl(`${GOOGLE_FILES_BASE}/v1beta/${fileName2}`, {
       method: "DELETE",
       headers: { "x-goog-api-key": apiKey }
     });
@@ -12573,7 +13162,7 @@ var OMNI_POLL_INTERVAL_MS = 1e4;
 var RESULT_TIMEOUT_MS3 = 15 * 60 * 1e3;
 var OMNI_INLINE_LIMIT_BYTES = 4 * 1024 * 1024;
 var MEDIA_PART_INLINE_LIMIT_BYTES = 10 * 1024 * 1024;
-var errorTextSchema = z22.string();
+var errorTextSchema = z23.string();
 function isGoogleGenaiVideoKnown(providerModelId) {
   return providerModelId.startsWith("veo-") || providerModelId.startsWith("gemini-omni");
 }
@@ -12768,7 +13357,7 @@ async function generateVeoVideo(req, ctx, createGenAIClient) {
     }
   };
   const jobId = operation.name ?? crypto.randomUUID();
-  const usage3 = videoUsage2(ctx, duration, req.resolution ?? "720p");
+  const usage4 = videoUsage2(ctx, duration, req.resolution ?? "720p");
   return {
     jobId,
     async poll() {
@@ -12806,7 +13395,7 @@ async function generateVeoVideo(req, ctx, createGenAIClient) {
           modelId
         );
       }
-      return { video: await downloadVideo(ctx, uri, apiKey, signal), usage: usage3, raw: finished };
+      return { video: await downloadVideo(ctx, uri, apiKey, signal), usage: usage4, raw: finished };
     }
   };
 }
@@ -12917,7 +13506,7 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
     return file.fileUri;
   };
   const cleanupUploads = async () => {
-    for (const fileName of uploadedNames) await deleteGoogleFile(fileName, apiKey, ctx.fetch);
+    for (const fileName2 of uploadedNames) await deleteGoogleFile(fileName2, apiKey, ctx.fetch);
   };
   const callCreate = (input) => client2.interactions.create(
     previousId !== void 0 ? {
@@ -13011,7 +13600,7 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
     if (!(extend && !chained)) await cleanupUploads();
   }
   const jobId = interaction.id ?? crypto.randomUUID();
-  const usage3 = videoUsage2(ctx, duration, omniResolution ?? "720p");
+  const usage4 = videoUsage2(ctx, duration, omniResolution ?? "720p");
   const getInteraction = async () => {
     try {
       const getParams = { stream: false };
@@ -13059,7 +13648,7 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
             bytes: Buffer.from(video.data, "base64"),
             mime: video.mime_type ?? "video/mp4"
           },
-          usage: usage3,
+          usage: usage4,
           raw: done,
           providerRef
         };
@@ -13067,7 +13656,7 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
       if (video.uri)
         return {
           video: await downloadVideo(ctx, video.uri, apiKey, signal),
-          usage: usage3,
+          usage: usage4,
           raw: done,
           providerRef
         };
@@ -13237,15 +13826,15 @@ async function prepareGoogleMediaPart(ref, mediaType, ctx) {
 // src/adapters/google-image.ts
 import { createGoogleGenerativeAI as createGoogleGenerativeAI2 } from "@ai-sdk/google";
 import { generateText } from "ai";
-import { z as z23 } from "zod";
+import { z as z24 } from "zod";
 var GEMINI_IMAGE_TIMEOUT_MS = 6e5;
 var defaultGoogleImageDeps = {
   createGoogleClient: (opts) => createGoogleGenerativeAI2(opts),
   runTextGeneration: (opts) => generateText(opts)
 };
-var causeChainNodeSchema = z23.object({
-  message: z23.string().optional(),
-  cause: z23.unknown().optional()
+var causeChainNodeSchema = z24.object({
+  message: z24.string().optional(),
+  cause: z24.unknown().optional()
 });
 function extractBlockReason(cause) {
   let current = cause;
@@ -13345,7 +13934,7 @@ function createGoogleImageAdapter(deps = defaultGoogleImageDeps) {
           retryable: false
         });
       }
-      const usage3 = {
+      const usage4 = {
         model: ctx.model.id,
         provider: "google-image",
         modality: "IMAGE",
@@ -13353,19 +13942,19 @@ function createGoogleImageAdapter(deps = defaultGoogleImageDeps) {
         quantity: images.length,
         imageQuality: "auto"
       };
-      if (req.size !== void 0) usage3.imageSize = req.size;
+      if (req.size !== void 0) usage4.imageSize = req.size;
       if (result.usage?.inputTokens !== void 0)
-        usage3.inputTokens = result.usage.inputTokens;
+        usage4.inputTokens = result.usage.inputTokens;
       if (result.usage?.outputTokens !== void 0)
-        usage3.outputTokens = result.usage.outputTokens;
-      return { images, usage: usage3, raw: result };
+        usage4.outputTokens = result.usage.outputTokens;
+      return { images, usage: usage4, raw: result };
     }
   };
 }
 var googleImageAdapter = createGoogleImageAdapter();
 
 // src/adapters/higgsfield.ts
-import { z as z24 } from "zod";
+import { z as z25 } from "zod";
 
 // src/adapters/higgsfield-motions.ts
 var MOTION_MAP = {
@@ -13584,14 +14173,14 @@ function klingEndpointFields(providerModelId) {
   const path = (providerModelId.startsWith("/") ? providerModelId.slice(1) : providerModelId).toLowerCase();
   return KLING_ENDPOINT_FIELDS.get(path) ?? UNLISTED_KLING_FIELDS;
 }
-var klingCreateSchema = z24.object({ status_url: z24.string() });
-var klingStatusSchema = z24.object({
-  status: z24.string().optional(),
-  video: z24.object({ url: z24.string().optional() }).optional(),
+var klingCreateSchema = z25.object({ status_url: z25.string() });
+var klingStatusSchema = z25.object({
+  status: z25.string().optional(),
+  video: z25.object({ url: z25.string().optional() }).optional(),
   error: JsonValueSchema.optional()
 });
-var errorMessageSchema = z24.object({ message: z24.string().optional() });
-var textValueSchema = z24.string();
+var errorMessageSchema = z25.object({ message: z25.string().optional() });
+var textValueSchema = z25.string();
 function klingErrorDetail(error) {
   if (error === void 0) return "unknown_error";
   const asText = textValueSchema.safeParse(error);
@@ -13747,8 +14336,8 @@ async function generateKlingVideo(req, ctx) {
     }
   };
 }
-var dopMotionOptionsSchema = z24.object({
-  modelkit: z24.object({ motion: z24.string().optional() }).optional()
+var dopMotionOptionsSchema = z25.object({
+  modelkit: z25.object({ motion: z25.string().optional() }).optional()
 });
 function dopMotionId(ctx) {
   const parsed = dopMotionOptionsSchema.safeParse(ctx.model.providerOptions);
@@ -13764,7 +14353,7 @@ function dopMotionId(ctx) {
     });
   return id;
 }
-var jsonRecordSchema = z24.record(z24.string(), JsonValueSchema);
+var jsonRecordSchema = z25.record(z25.string(), JsonValueSchema);
 var DOP_URL_KEYS = ["url", "video_url", "output_url", "result_url", "download_url"];
 var DOP_NESTED_KEYS = ["video", "result", "results"];
 function dopUrl(body) {
@@ -13792,8 +14381,8 @@ function dopUrl(body) {
   }
   return void 0;
 }
-var dopCreateSchema = z24.object({ id: z24.string() });
-var dopStatusSchema = z24.object({ status: z24.string().optional() });
+var dopCreateSchema = z25.object({ id: z25.string() });
+var dopStatusSchema = z25.object({ status: z25.string().optional() });
 async function generateDopVideo(req, ctx) {
   const firstFrame = req.firstFrame ? refToUrl(req.firstFrame) : void 0;
   if (!firstFrame) {
@@ -13920,7 +14509,7 @@ var higgsfieldAdapter = {
 };
 
 // src/adapters/luma.ts
-import { z as z25 } from "zod";
+import { z as z26 } from "zod";
 
 // src/adapters/mp4-duration.ts
 function mp4DurationSeconds(bytes) {
@@ -13976,7 +14565,7 @@ function findBox(bytes, from, to, type) {
 }
 
 // src/adapters/luma.ts
-var DEFAULT_BASE_URL = "https://agents.lumalabs.ai";
+var DEFAULT_BASE_URL2 = "https://agents.lumalabs.ai";
 var IMAGE_POLL_INTERVAL_MS = 3e3;
 var IMAGE_TIMEOUT_MS = 5 * 6e4;
 var VIDEO_POLL_INTERVAL_MS = 5e3;
@@ -14005,7 +14594,7 @@ function lumaModel(ctx) {
   return ctx.model.providerModelId.toLowerCase();
 }
 function lumaBase(ctx) {
-  return ctx.endpoints?.luma ?? DEFAULT_BASE_URL;
+  return ctx.endpoints?.luma ?? DEFAULT_BASE_URL2;
 }
 function lumaKey(ctx) {
   const key = ctx.keys.luma;
@@ -14137,17 +14726,17 @@ function lumaReframeResolution(req, ctx) {
   });
   return "720p";
 }
-var lumaGenerationSchema = z25.looseObject({
-  id: z25.string(),
-  state: z25.string().optional(),
-  output: z25.array(
-    z25.looseObject({
-      type: z25.string().optional(),
-      url: z25.string().optional()
+var lumaGenerationSchema = z26.looseObject({
+  id: z26.string(),
+  state: z26.string().optional(),
+  output: z26.array(
+    z26.looseObject({
+      type: z26.string().optional(),
+      url: z26.string().optional()
     })
   ).optional(),
-  failure_reason: z25.string().optional().nullable(),
-  failure_code: z25.string().optional().nullable()
+  failure_reason: z26.string().optional().nullable(),
+  failure_code: z26.string().optional().nullable()
 });
 function lumaFailure(ctx, code, reason) {
   const detail = reason ?? code ?? "unknown_error";
@@ -14579,7 +15168,7 @@ var lumaAdapter = {
 
 // src/adapters/openai-image.ts
 import OpenAI from "openai";
-import { z as z26 } from "zod";
+import { z as z27 } from "zod";
 var MAX_EDIT_REFERENCES = 16;
 function aspectRatioToSize(ratio, tier) {
   const TIER_LONG_EDGE = {
@@ -14662,9 +15251,9 @@ function toProviderError2(cause, ctx) {
     ctx.model.id
   );
 }
-var openaiUsageSchema = z26.object({
-  input_tokens: z26.number().optional(),
-  output_tokens: z26.number().optional()
+var openaiUsageSchema = z27.object({
+  input_tokens: z27.number().optional(),
+  output_tokens: z27.number().optional()
 });
 var openaiImageAdapter = {
   id: "openai-image",
@@ -14789,7 +15378,7 @@ var openaiImageAdapter = {
     const usageParsed = openaiUsageSchema.safeParse(response.usage);
     const inputTokens = usageParsed.success ? usageParsed.data.input_tokens : void 0;
     const outputTokens = usageParsed.success ? usageParsed.data.output_tokens : void 0;
-    const usage3 = {
+    const usage4 = {
       model: ctx.model.id,
       provider: "openai-image",
       modality: "IMAGE",
@@ -14797,10 +15386,10 @@ var openaiImageAdapter = {
       quantity: images.length,
       imageQuality: quality
     };
-    if (inputTokens !== void 0) usage3.inputTokens = inputTokens;
-    if (outputTokens !== void 0) usage3.outputTokens = outputTokens;
-    if (req.size !== void 0) usage3.imageSize = req.size;
-    return { images, usage: usage3, raw: response };
+    if (inputTokens !== void 0) usage4.inputTokens = inputTokens;
+    if (outputTokens !== void 0) usage4.outputTokens = outputTokens;
+    if (req.size !== void 0) usage4.imageSize = req.size;
+    return { images, usage: usage4, raw: response };
   }
 };
 
@@ -15056,16 +15645,16 @@ async function generateProdiaImage(req, ctx) {
     new Uint8Array(await res.arrayBuffer()),
     res.headers.get("content-type")
   );
-  const usage3 = {
+  const usage4 = {
     model: ctx.model.id,
     provider: "prodia",
     modality: "IMAGE",
     unit: "images",
     quantity: 1
   };
-  if (req.size !== void 0) usage3.imageSize = req.size;
-  if (req.quality !== void 0) usage3.imageQuality = req.quality;
-  return { images: [{ bytes, mime }], usage: usage3, raw: void 0 };
+  if (req.size !== void 0) usage4.imageSize = req.size;
+  if (req.quality !== void 0) usage4.imageQuality = req.quality;
+  return { images: [{ bytes, mime }], usage: usage4, raw: void 0 };
 }
 var prodiaAdapter = {
   id: "prodia",
@@ -15076,13 +15665,13 @@ var prodiaAdapter = {
 import Replicate from "replicate";
 
 // src/adapters/replicate-http.ts
-import { z as z27 } from "zod";
-var predictionSchema = z27.object({
-  id: z27.string(),
-  status: z27.string(),
+import { z as z28 } from "zod";
+var predictionSchema = z28.object({
+  id: z28.string(),
+  status: z28.string(),
   output: JsonValueSchema.optional(),
   error: JsonValueSchema.optional(),
-  metrics: z27.object({ predict_time: z27.number().optional() }).optional()
+  metrics: z28.object({ predict_time: z28.number().optional() }).optional()
 });
 function createReplicateHttpClient(opts) {
   async function request(path, body, signal) {
@@ -15133,7 +15722,7 @@ function createReplicateHttpClient(opts) {
 }
 
 // src/adapters/replicate.ts
-import { z as z28 } from "zod";
+import { z as z29 } from "zod";
 var POLL_INTERVAL_MS = 1e3;
 var POLL_TIMEOUT_MS = 15 * 6e4;
 var defaultReplicateDeps = {
@@ -15173,8 +15762,8 @@ function route(providerModelId) {
   if (!providerModelId.includes("/")) return { version: providerModelId };
   return { model: providerModelId };
 }
-var replicateResponseStatusSchema = z28.object({
-  response: z28.object({ status: z28.number() }).partial()
+var replicateResponseStatusSchema = z29.object({
+  response: z29.object({ status: z29.number() }).partial()
 });
 function replicateError(cause, ctx) {
   if (isModelKitError(cause)) return cause;
@@ -15197,7 +15786,7 @@ function isUrlFn(value) {
 function extractOutputUrl(output) {
   const root = Array.isArray(output) ? output[0] : output;
   if (root === void 0) throw noOutput();
-  const asString = z28.string().safeParse(root);
+  const asString = z29.string().safeParse(root);
   if (asString.success && asString.data) return asString.data;
   if (root instanceof URL) return root.href;
   if (isObjectOutput(root)) {
@@ -15206,15 +15795,15 @@ function extractOutputUrl(output) {
       if (urlValue !== void 0 && isUrlFn(urlValue)) {
         const called = urlValue();
         if (called instanceof URL) return called.href;
-        const calledString = z28.string().safeParse(called);
+        const calledString = z29.string().safeParse(called);
         if (calledString.success && calledString.data) return calledString.data;
       } else {
-        const direct = z28.string().safeParse(urlValue);
+        const direct = z29.string().safeParse(urlValue);
         if (direct.success && direct.data) return direct.data;
       }
     }
     if ("default" in root) {
-      const fallback = z28.string().safeParse(root.default);
+      const fallback = z29.string().safeParse(root.default);
       if (fallback.success && fallback.data) return fallback.data;
     }
   }
@@ -15275,7 +15864,7 @@ async function runPrediction(ctx, input, deps) {
     }
     return void 0;
   }, pollOpts);
-  const predictTime = z28.number().safeParse(done.metrics?.predict_time);
+  const predictTime = z29.number().safeParse(done.metrics?.predict_time);
   if (done.output === void 0) throw noOutput();
   return {
     url: extractOutputUrl(done.output),
@@ -15521,9 +16110,9 @@ async function generateReplicateImage(req, ctx, deps) {
   }
   const { url, predictTime, raw } = await runPrediction(ctx, input, deps);
   const file = await download(ctx, url);
-  const usage3 = imageUsage3(ctx, req);
-  usage3.quantity = predictTime;
-  return { images: [{ bytes: file.bytes, mime: file.mime }], usage: usage3, raw, outputUrl: url };
+  const usage4 = imageUsage3(ctx, req);
+  usage4.quantity = predictTime;
+  return { images: [{ bytes: file.bytes, mime: file.mime }], usage: usage4, raw, outputUrl: url };
 }
 function klingInput(req) {
   const images = (req.firstFrame ? [req.firstFrame] : []).map(refToUrl);
@@ -15628,7 +16217,7 @@ async function generateReplicateVideo(req, ctx, deps) {
       throw replicateError(cause, ctx);
     }
     if (current.status === "succeeded") {
-      const predictTime = z28.number().safeParse(current.metrics?.predict_time);
+      const predictTime = z29.number().safeParse(current.metrics?.predict_time);
       if (current.output === void 0) throw noOutput();
       return {
         url: extractOutputUrl(current.output),
@@ -15647,9 +16236,9 @@ async function generateReplicateVideo(req, ctx, deps) {
     return void 0;
   };
   const toResult = (done) => {
-    const usage3 = videoUsage5(ctx, req);
-    usage3.quantity = done.predictTime;
-    return { video: { url: done.url }, usage: usage3, raw: done.raw };
+    const usage4 = videoUsage5(ctx, req);
+    usage4.quantity = done.predictTime;
+    return { video: { url: done.url }, usage: usage4, raw: done.raw };
   };
   return {
     jobId,
@@ -15756,21 +16345,21 @@ async function generateReplicateAudio(req, ctx, deps) {
     input = specResult.input;
   }
   const { url, predictTime, raw } = await runPrediction(ctx, input, deps);
-  const usage3 = audioUsage(ctx, req);
+  const usage4 = audioUsage(ctx, req);
   if (ctx.model.usage.unit === "seconds") {
     const seconds3 = elevenLabsMusicDuration(req.durationSeconds);
-    usage3.quantity = seconds3;
-    usage3.durationSeconds = seconds3;
+    usage4.quantity = seconds3;
+    usage4.durationSeconds = seconds3;
   } else {
-    usage3.quantity = predictTime;
+    usage4.quantity = predictTime;
   }
-  return { audio: { url }, usage: usage3, raw: predictionRaw(raw) };
+  return { audio: { url }, usage: usage4, raw: predictionRaw(raw) };
 }
 
 // src/adapters/xai-video.ts
 import { experimental_generateVideo } from "ai";
 import { createXai as createXai2 } from "@ai-sdk/xai";
-import { z as z29 } from "zod";
+import { z as z30 } from "zod";
 var POLL_INTERVAL_MS2 = 1e4;
 var POLL_TIMEOUT_MS2 = 15 * 6e4;
 var defaultXaiDeps = {
@@ -15783,8 +16372,8 @@ function createXaiVideoAdapter(deps = defaultXaiDeps) {
   };
 }
 var xaiVideoAdapter = createXaiVideoAdapter();
-var xaiResponseStatusSchema = z29.object({
-  response: z29.object({ status: z29.number() }).partial()
+var xaiResponseStatusSchema = z30.object({
+  response: z30.object({ status: z30.number() }).partial()
 });
 function xaiError(cause, ctx) {
   const info = describeThrown(cause);
@@ -15792,12 +16381,12 @@ function xaiError(cause, ctx) {
   const status = info.status ?? (nested.success ? nested.data.response?.status : void 0);
   return classifyThrown("xai-video", cause, ctx.model.id, { status, body: info.body });
 }
-var xaiMetaSchema = z29.object({
+var xaiMetaSchema = z30.object({
   videoUrl: JsonValueSchema.optional(),
   costInUsdTicks: JsonValueSchema.optional(),
   duration: JsonValueSchema.optional()
 });
-var xaiProviderMetadataSchema = z29.object({ xai: xaiMetaSchema });
+var xaiProviderMetadataSchema = z30.object({ xai: xaiMetaSchema });
 function metaOf(result) {
   const parsed = xaiProviderMetadataSchema.partial().safeParse(result.providerMetadata);
   if (!parsed.success) return {};
@@ -15909,12 +16498,12 @@ async function mapResult(ctx, req, result) {
   if (first) {
     return {
       video: { bytes: Uint8Array.from(first.uint8Array), mime: first.mediaType },
-      usage: usage2(ctx, req, metaOf(result)),
+      usage: usage3(ctx, req, metaOf(result)),
       raw: result.providerMetadata
     };
   }
   const meta = metaOf(result);
-  const videoUrl = z29.string().safeParse(meta.videoUrl);
+  const videoUrl = z30.string().safeParse(meta.videoUrl);
   if (videoUrl.success && videoUrl.data) {
     const res = await ctx.fetch(videoUrl.data);
     if (!res.ok) {
@@ -15924,7 +16513,7 @@ async function mapResult(ctx, req, result) {
     const mime = res.headers.get("content-type") ?? "video/mp4";
     return {
       video: { bytes: new Uint8Array(await res.arrayBuffer()), mime },
-      usage: usage2(ctx, req, meta),
+      usage: usage3(ctx, req, meta),
       raw: result.providerMetadata
     };
   }
@@ -15936,8 +16525,8 @@ async function mapResult(ctx, req, result) {
     retryable: false
   });
 }
-function usage2(ctx, req, meta) {
-  const costTicks = z29.number().safeParse(meta.costInUsdTicks);
+function usage3(ctx, req, meta) {
+  const costTicks = z30.number().safeParse(meta.costInUsdTicks);
   const cost = costTicks.success ? costTicks.data / 1e10 : void 0;
   const event = {
     model: ctx.model.id,
@@ -15976,7 +16565,8 @@ var DEFAULT_ADAPTERS = {
   prodia: prodiaAdapter,
   higgsfield: higgsfieldAdapter,
   "xai-video": xaiVideoAdapter,
-  luma: lumaAdapter
+  luma: lumaAdapter,
+  fish: fishAdapter
 };
 
 // src/catalog/data-only.ts
@@ -16000,6 +16590,8 @@ function adapterServesType(adapter, type) {
   if (type === "dubbing") return adapter.dub !== void 0;
   if (type === "lipsync") return adapter.lipsync !== void 0;
   if (type === "transcription") return adapter.transcribe !== void 0;
+  if (type === "voice-clone") return adapter.createVoice !== void 0;
+  if (type === "voice-design") return adapter.designVoice !== void 0;
   if (type === "video-tool") return adapter.submitVideoTool !== void 0;
   if (type === "model3d") return adapter.submitModel3d !== void 0;
   return adapter.languageModel !== void 0;
@@ -16013,12 +16605,15 @@ var FAL_NODE_TYPES = /* @__PURE__ */ new Set([
   "transcription",
   "mask"
 ]);
+function fishSpeech(row2) {
+  return row2.adapter === "fish" && row2.type === "speech";
+}
 function dataOnlyCapable(row2) {
   const adapter = DEFAULT_ADAPTERS[row2.adapter];
   if (!adapterServesType(adapter, row2.type)) {
     return { ok: false, reason: `${row2.adapter} cannot serve ${row2.type} rows` };
   }
-  if (FAL_NODE_TYPES.has(row2.type) && row2.adapter !== "fal") {
+  if (FAL_NODE_TYPES.has(row2.type) && row2.adapter !== "fal" && !fishSpeech(row2)) {
     return { ok: false, reason: `only fal serves ${row2.type} rows` };
   }
   switch (row2.adapter) {
@@ -16096,6 +16691,10 @@ function dataOnlyCapable(row2) {
       const reason = lumaRowRefusal(row2);
       return reason === void 0 ? { ok: true } : { ok: false, reason };
     }
+    case "fish": {
+      const reason = fishRowRefusal(row2);
+      return reason === void 0 ? { ok: true } : { ok: false, reason };
+    }
     default: {
       return assertNever(row2.adapter);
     }
@@ -16115,22 +16714,22 @@ var liveState = store[KEY] ??= {
 var liveCatalog = liveState.catalog;
 
 // src/catalog/load.ts
-import { z as z30 } from "zod";
+import { z as z31 } from "zod";
 function rowCount() {
   return liveCatalog.list({ lifecycle: ["active", "deprecated", "retired"] }).length;
 }
 function authHeaders(auth) {
   return "bearer" in auth ? { Authorization: `Bearer ${auth.bearer}` } : { "x-api-key": auth.apiKey };
 }
-var EnvelopeSchema = z30.object({
-  rows: z30.array(JsonValueSchema),
-  revision: z30.string().optional()
+var EnvelopeSchema = z31.object({
+  rows: z31.array(JsonValueSchema),
+  revision: z31.string().optional()
 });
 function dropReason(path, message) {
   return path.length === 0 ? message : `${path.map((part) => String(part)).join(".")} ${message}`;
 }
 function rowId(raw) {
-  const parsed = z30.object({ id: z30.string().min(1) }).loose().safeParse(raw);
+  const parsed = z31.object({ id: z31.string().min(1) }).loose().safeParse(raw);
   if (!parsed.success) return "?";
   return parsed.data.id;
 }
@@ -17046,8 +17645,17 @@ 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,
     {
@@ -17169,6 +17777,106 @@ 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/keys.ts
 var ADAPTER_KEYS = {
   "ai-sdk-text": [],
@@ -17181,7 +17889,8 @@ var ADAPTER_KEYS = {
   prodia: ["prodia"],
   higgsfield: ["higgsfield"],
   "xai-video": ["xai"],
-  luma: ["luma"]
+  luma: ["luma"],
+  fish: ["fish"]
 };
 function hasAll(keys, names) {
   if (!keys) return false;
@@ -17240,6 +17949,7 @@ function keysFromEnv(env, prefix = "") {
   set("higgsfield", pick2("HIGGSFIELD_API_KEY"));
   set("higgsfieldSecret", pick2("HIGGSFIELD_API_SECRET"));
   set("luma", pick2("LUMA_API_KEY", "LUMA_AGENTS_API_KEY"));
+  set("fish", pick2("FISH_AUDIO_API_KEY"));
   set("cloudflare", pick2("CLOUDFLARE_AI_GATEWAY_TOKEN", "CF_AIG_TOKEN"));
   const bearer = pick2("AWS_BEARER_TOKEN_BEDROCK");
   const accessKeyId = pick2("AWS_ACCESS_KEY_ID");
@@ -17262,7 +17972,7 @@ function keysFromEnv(env, prefix = "") {
 
 // src/text.ts
 import "ai";
-import { z as z31 } from "zod";
+import { z as z32 } from "zod";
 function mergeTextProviderOptions(catalog, caller) {
   const merged = {};
   const catalogOptions = catalog.providerOptions ?? {};
@@ -17301,17 +18011,17 @@ function passthroughTextModel(model) {
     pricingKind: "token"
   };
 }
-var perplexityCostSchema = z31.object({
-  perplexity: z31.object({ cost: z31.object({ totalCost: z31.number().finite().nonnegative() }) })
+var perplexityCostSchema = z32.object({
+  perplexity: z32.object({ cost: z32.object({ totalCost: z32.number().finite().nonnegative() }) })
 });
 function reportedCostUsd(metadata) {
   const parsed = perplexityCostSchema.safeParse(metadata);
   return parsed.success ? parsed.data.perplexity.cost.totalCost : void 0;
 }
-function textUsageEvent(model, provider, usage3, providerMetadata) {
-  const inputTokens = usage3.inputTokens ?? 0;
-  const outputTokens = usage3.outputTokens ?? 0;
-  const total = usage3.totalTokens ?? inputTokens + outputTokens;
+function textUsageEvent(model, provider, usage4, providerMetadata) {
+  const inputTokens = usage4.inputTokens ?? 0;
+  const outputTokens = usage4.outputTokens ?? 0;
+  const total = usage4.totalTokens ?? inputTokens + outputTokens;
   const event = {
     model: model.id,
     provider,
@@ -17346,7 +18056,7 @@ function createTextSteps(initial, resolve, callerProviderOptions) {
       return { ...prepared, providerOptions };
     };
   }
-  function usage3(results) {
+  function usage4(results) {
     const grouped = /* @__PURE__ */ new Map();
     for (const [index, step] of results.entries()) {
       const target = steps.get(index) ?? initial;
@@ -17370,7 +18080,7 @@ function createTextSteps(initial, resolve, callerProviderOptions) {
     }
     return [...grouped.values()];
   }
-  return { wrap, usage: usage3 };
+  return { wrap, usage: usage4 };
 }
 
 // src/client.ts
@@ -17382,7 +18092,7 @@ import {
   streamObject as aiStreamObject2,
   streamText as aiStreamText2
 } from "ai";
-import { z as z32 } from "zod";
+import { z as z33 } from "zod";
 
 // src/generated-image.ts
 function toGeneratedImage(image) {
@@ -17504,9 +18214,9 @@ function providerJobMethods(kind, deps) {
       if (!methods.result) throw missing3(adapter, ctx.model);
       const result = await methods.result(sub, ctx, webhook);
       const base = { ...result.usage, model: ctx.model.id, unit: ctx.model.usage.unit };
-      const usage3 = sub.submittedAt === void 0 ? base : deps.timed(base, sub.submittedAt);
-      await deps.emit(usage3);
-      return { ...result, usage: usage3 };
+      const usage4 = sub.submittedAt === void 0 ? base : deps.timed(base, sub.submittedAt);
+      await deps.emit(usage4);
+      return { ...result, usage: usage4 };
     },
     async cancel(sub, call = {}) {
       const { methods, adapter, ctx } = target(sub.model, call);
@@ -17531,6 +18241,9 @@ 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.
@@ -17545,6 +18258,8 @@ 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) {
@@ -17587,6 +18302,14 @@ 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);
+}
 
 // src/client-audio.ts
 function noMethod(adapter, model, what) {
@@ -17676,6 +18399,49 @@ function audioNodeMethods(deps) {
       const result = await adapter.transcribe(resolved, ctx);
       return { ...result, usage: await deps.settle(model, result.usage, startedAt) };
     },
+    async createVoice(req, call = {}) {
+      mediaRetryOptions(call);
+      const { abortSignal, ...request } = req;
+      const model = catalog.require(req.model);
+      const { request: resolved, warnings } = resolveVoiceCloneCall(model, request);
+      const { adapter, ctx } = prepare(model, call, abortSignal, warnings);
+      if (!adapter.createVoice) throw noMethod(adapter, model, "clone a voice");
+      const startedAt = Date.now();
+      const result = await adapter.createVoice(resolved, ctx);
+      return { ...result, usage: await deps.settle(model, result.usage, startedAt) };
+    },
+    async designVoice(req, call = {}) {
+      mediaRetryOptions(call);
+      const { abortSignal, ...request } = req;
+      const model = catalog.require(req.model);
+      const { request: resolved, warnings } = resolveVoiceDesignCall(
+        model,
+        request,
+        call.onUnsupported ?? "error"
+      );
+      const { adapter, ctx } = prepare(model, call, abortSignal, warnings);
+      if (!adapter.designVoice) throw noMethod(adapter, model, "design a voice");
+      const startedAt = Date.now();
+      const result = await adapter.designVoice(resolved, ctx);
+      return { ...result, usage: await deps.settle(model, result.usage, startedAt) };
+    },
+    async deleteVoice(req, call = {}) {
+      mediaRetryOptions(call);
+      const { abortSignal, ...request } = req;
+      const model = catalog.require(req.model);
+      assertMethod(model, "createVoice");
+      const voiceId = request.voiceId.trim();
+      if (!voiceId) {
+        throw new ModelKitError({
+          code: "missing_input",
+          message: `${model.id} requires voiceId`,
+          model: model.id
+        });
+      }
+      const { adapter, ctx } = prepare(model, call, abortSignal, []);
+      if (!adapter.deleteVoice) throw noMethod(adapter, model, "delete a voice");
+      await adapter.deleteVoice({ model: model.id, voiceId }, ctx);
+    },
     async dub(req, call = {}) {
       mediaRetryOptions(call);
       const { abortSignal, ...request } = req;
@@ -17690,10 +18456,10 @@ function audioNodeMethods(deps) {
       const startedAt = Date.now();
       const outcome = await adapter.dub(resolved, ctx);
       if (outcome.kind === "audio") {
-        const { audio, usage: usage3, raw } = outcome.result;
+        const { audio, usage: usage4, raw } = outcome.result;
         return {
           audio,
-          usage: await deps.settle(model, usage3, startedAt),
+          usage: await deps.settle(model, usage4, startedAt),
           ...raw !== void 0 && { raw }
         };
       }
@@ -17731,10 +18497,10 @@ function textAbortOptions(sdk, context) {
   if (abortSignal) return { abortSignal };
   return {};
 }
-var catalogId = z32.string();
-var exactSize = z32.templateLiteral([z32.number(), "x", z32.number()]);
-var sdkRatio = z32.templateLiteral([z32.number(), ":", z32.number()]);
-var sizeTier = z32.enum(["512", "1K", "2K", "4K"]);
+var catalogId = z33.string();
+var exactSize = z33.templateLiteral([z33.number(), "x", z33.number()]);
+var sdkRatio = z33.templateLiteral([z33.number(), ":", z33.number()]);
+var sizeTier = z33.enum(["512", "1K", "2K", "4K"]);
 function isCatalogId(model) {
   return catalogId.safeParse(model).success;
 }
@@ -17947,7 +18713,7 @@ function createModelKit(config, adapters) {
     const images = result.images.map(
       (file) => toGeneratedImage({ bytes: file.uint8Array, mime: file.mediaType })
     );
-    const usage3 = timed(
+    const usage4 = timed(
       {
         model: model.modelId,
         provider: model.provider,
@@ -17957,8 +18723,8 @@ function createModelKit(config, adapters) {
       },
       startedAt
     );
-    await emit(usage3);
-    return { image: images[0], images, usage: usage3, raw: result, warnings };
+    await emit(usage4);
+    return { image: images[0], images, usage: usage4, raw: result, warnings };
   }
   function passthroughVideo(req, call) {
     const {
@@ -18034,7 +18800,7 @@ function createModelKit(config, adapters) {
     let settled;
     let failed = false;
     const done = pending.then(async (result) => {
-      const usage3 = timed(
+      const usage4 = timed(
         {
           model: model.modelId,
           provider: model.provider,
@@ -18047,10 +18813,10 @@ function createModelKit(config, adapters) {
       );
       settled = {
         video: { bytes: result.video.uint8Array, mime: result.video.mediaType },
-        usage: usage3,
+        usage: usage4,
         raw: result
       };
-      await emit(usage3);
+      await emit(usage4);
       return settled;
     });
     done.catch(() => {
@@ -18077,12 +18843,12 @@ function createModelKit(config, adapters) {
         if (settled) return settled;
         const pending = result(opts).then(
           async (out) => {
-            const usage3 = timed(
+            const usage4 = timed(
               { ...out.usage, model: model.id, unit: model.usage.unit },
               startedAt
             );
-            await emit(usage3);
-            return { ...out, usage: usage3 };
+            await emit(usage4);
+            return { ...out, usage: usage4 };
           },
           (cause) => {
             settled = void 0;
@@ -18102,9 +18868,9 @@ function createModelKit(config, adapters) {
       return context(model, signal ? { ...call, signal } : call, warnings);
     },
     async settle(model, event, startedAt) {
-      const usage3 = timed({ ...event, model: model.id, unit: model.usage.unit }, startedAt);
-      await emit(usage3);
-      return usage3;
+      const usage4 = timed({ ...event, model: model.id, unit: model.usage.unit }, startedAt);
+      await emit(usage4);
+      return usage4;
     },
     wrapVideoJob
   });
@@ -18140,11 +18906,11 @@ function createModelKit(config, adapters) {
         });
       const startedAt = Date.now();
       const result = await adapter.generateImage(request, ctx);
-      const usage3 = timed(
+      const usage4 = timed(
         { ...result.usage, model: model.id, unit: model.usage.unit },
         startedAt
       );
-      await emit(usage3);
+      await emit(usage4);
       const images = result.images.map(toGeneratedImage);
       const first = images[0];
       if (!first)
@@ -18153,7 +18919,7 @@ function createModelKit(config, adapters) {
           message: `${model.id} returned no image`,
           model: model.id
         });
-      return { ...result, usage: usage3, image: first, images, warnings };
+      return { ...result, usage: usage4, image: first, images, warnings };
     },
     async generateVideo(req, call = {}) {
       mediaRetryOptions(call);
@@ -18252,12 +19018,12 @@ function createModelKit(config, adapters) {
       }
       const startedAt = Date.now();
       const result = await adapter.generateAudio(resolved, ctx);
-      const usage3 = timed(
+      const usage4 = timed(
         { ...result.usage, model: model.id, unit: model.usage.unit },
         startedAt
       );
-      await emit(usage3);
-      return { ...result, usage: usage3 };
+      await emit(usage4);
+      return { ...result, usage: usage4 };
     },
     async prepareMediaPart(modelId, ref, mediaType, call = {}) {
       mediaRetryOptions(call);
@@ -18337,8 +19103,8 @@ function createModelKit(config, adapters) {
           options.prepareStep ?? options.experimental_prepareStep
         ),
         onFinish: async (event) => {
-          for (const usage3 of steps.usage(event.steps))
-            await emit(textTimed(withRowUnit(usage3), event.steps.length, startedAt));
+          for (const usage4 of steps.usage(event.steps))
+            await emit(textTimed(withRowUnit(usage4), event.steps.length, startedAt));
           await callerOnFinish?.(event);
         }
       });
@@ -18460,6 +19226,7 @@ export {
   falAdapter,
   falWebhookBodySchema,
   fileCatalogCache,
+  fishAdapter,
   generateAudio,
   generateImage,
   generateObject,
@@ -18512,6 +19279,8 @@ export {
   resolveTranscriptionOptions,
   resolveVideoOptions,
   resolveVoiceChangeOptions,
+  resolveVoiceCloneOptions,
+  resolveVoiceDesignOptions,
   shouldRetry,
   sleep,
   sortCatalogRows,