@tryhamster/gerbil 1.2.2 → 1.3.0

package/dist/defaults-CudhglS1.mjs

@@ -0,0 +1,47 @@
+//#region src/gpu/defaults.ts
+/**
+* Default model per capability. Kept in its own tiny module (no heavy imports)
+* so the React hooks can resolve defaults without statically pulling in the GPU
+* engine — they import the engine dynamically to stay light.
+*/
+const DEFAULT_MODELS = {
+	text: "mlx-community/Qwen3.5-0.8B-4bit",
+	vision: "mlx-community/Qwen3.5-0.8B-4bit",
+	embedding: "mlx-community/embeddinggemma-300m-4bit",
+	tts: "nineninesix/kani-tts-450m-0.2-ft",
+	ttsOute: "OuteAI/OuteTTS-1.0-0.6B",
+	stt: "UsefulSensors/moonshine-base"
+};
+/**
+* OuteTTS auxiliary assets. The DAC.speech codec ships only `.pth` (non-safetensors),
+* so the folded weights (prep-dac-speech.py output) and the preset speaker JSONs are
+* hosted on the Cloudflare R2 model mirror (same public host the engine reads via
+* GERBIL_MODEL_BASE — range + CORS, free egress, no redirect). Override via the OuteTTS
+* options when self-hosting.
+*/
+const OUTETTS_ASSET_BASE = "https://pub-0522ecec25fc412bb35a523f3e3f63c6.r2.dev/tryhamster/gerbil-outetts-assets";
+const OUTETTS_ASSETS = {
+	codecWeightsURL: `${OUTETTS_ASSET_BASE}/dac-speech-folded.safetensors`,
+	speakersBaseURL: `${OUTETTS_ASSET_BASE}/speakers`,
+	defaultVoice: "en-female-1-neutral"
+};
+/** Bundled OuteTTS preset voices (the edwko/OuteTTS v3 default_speakers set). */
+const OUTETTS_PRESET_VOICES = [
+	"en-female-1-neutral",
+	"en-female-2-neutral",
+	"en-male-1-neutral",
+	"en-male-2-neutral",
+	"en-male-3-neutral",
+	"en-male-4-neutral"
+];
+/** Resolve the model repo for a set of options, falling back to the defaults. */
+function resolveDefaultRepo(opts) {
+	if (opts.repo) return opts.repo;
+	if (opts.embedding) return DEFAULT_MODELS.embedding;
+	if (opts.enableVision) return DEFAULT_MODELS.vision;
+	return DEFAULT_MODELS.text;
+}
+
+//#endregion
+export { resolveDefaultRepo as i, OUTETTS_ASSETS as n, OUTETTS_PRESET_VOICES as r, DEFAULT_MODELS as t };
+//# sourceMappingURL=defaults-CudhglS1.mjs.map

package/dist/defaults-CudhglS1.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"defaults-CudhglS1.mjs","names":[],"sources":["../src/gpu/defaults.ts"],"sourcesContent":["/**\n * Default model per capability. Kept in its own tiny module (no heavy imports)\n * so the React hooks can resolve defaults without statically pulling in the GPU\n * engine — they import the engine dynamically to stay light.\n */\nexport const DEFAULT_MODELS = {\n  /** Text generation (also the vision-capable checkpoint). */\n  text: \"mlx-community/Qwen3.5-0.8B-4bit\",\n  /** Image understanding — same checkpoint, vision tower built on demand. */\n  vision: \"mlx-community/Qwen3.5-0.8B-4bit\",\n  /** Text embeddings. */\n  embedding: \"mlx-community/embeddinggemma-300m-4bit\",\n  /** Text-to-speech (Apache-2.0; LFM2-350M backbone + NanoCodec). Small/fast default. */\n  tts: \"nineninesix/kani-tts-450m-0.2-ft\",\n  /**\n   * Selectable many-voices TTS (OuteTTS-1.0-0.6B: Qwen3-0.6B codec-LM + DAC.speech).\n   * Pass this as the `repo`/`model` to engine.speak()/useTTS to route to OuteTTS.\n   */\n  ttsOute: \"OuteAI/OuteTTS-1.0-0.6B\",\n  /** Speech-to-text. */\n  stt: \"UsefulSensors/moonshine-base\",\n} as const;\n\n/**\n * OuteTTS auxiliary assets. The DAC.speech codec ships only `.pth` (non-safetensors),\n * so the folded weights (prep-dac-speech.py output) and the preset speaker JSONs are\n * hosted on the Cloudflare R2 model mirror (same public host the engine reads via\n * GERBIL_MODEL_BASE — range + CORS, free egress, no redirect). Override via the OuteTTS\n * options when self-hosting.\n */\nconst OUTETTS_ASSET_BASE =\n  \"https://pub-0522ecec25fc412bb35a523f3e3f63c6.r2.dev/tryhamster/gerbil-outetts-assets\";\nexport const OUTETTS_ASSETS = {\n  /** Folded DAC.speech.v1.0 encoder+decoder weights (prep-dac-speech.py output). */\n  codecWeightsURL: `${OUTETTS_ASSET_BASE}/dac-speech-folded.safetensors`,\n  /** Base URL for the preset speaker JSONs (`{name}.json`). */\n  speakersBaseURL: `${OUTETTS_ASSET_BASE}/speakers`,\n  /** The default preset voice. */\n  defaultVoice: \"en-female-1-neutral\",\n} as const;\n\n/** Bundled OuteTTS preset voices (the edwko/OuteTTS v3 default_speakers set). */\nexport const OUTETTS_PRESET_VOICES = [\n  \"en-female-1-neutral\",\n  \"en-female-2-neutral\",\n  \"en-male-1-neutral\",\n  \"en-male-2-neutral\",\n  \"en-male-3-neutral\",\n  \"en-male-4-neutral\",\n] as const;\n\n/** Resolve the model repo for a set of options, falling back to the defaults. */\nexport function resolveDefaultRepo(opts: {\n  repo?: string;\n  embedding?: boolean;\n  enableVision?: boolean;\n}): string {\n  if (opts.repo) return opts.repo;\n  if (opts.embedding) return DEFAULT_MODELS.embedding;\n  if (opts.enableVision) return DEFAULT_MODELS.vision;\n  return DEFAULT_MODELS.text;\n}\n"],"mappings":";;;;;;AAKA,MAAa,iBAAiB;CAE5B,MAAM;CAEN,QAAQ;CAER,WAAW;CAEX,KAAK;CAKL,SAAS;CAET,KAAK;CACN;;;;;;;;AASD,MAAM,qBACJ;AACF,MAAa,iBAAiB;CAE5B,iBAAiB,GAAG,mBAAmB;CAEvC,iBAAiB,GAAG,mBAAmB;CAEvC,cAAc;CACf;;AAGD,MAAa,wBAAwB;CACnC;CACA;CACA;CACA;CACA;CACA;CACD;;AAGD,SAAgB,mBAAmB,MAIxB;AACT,KAAI,KAAK,KAAM,QAAO,KAAK;AAC3B,KAAI,KAAK,UAAW,QAAO,eAAe;AAC1C,KAAI,KAAK,aAAc,QAAO,eAAe;AAC7C,QAAO,eAAe"}

package/dist/frameworks/express.d.mts

@@ -1,4 +1,5 @@
-import { d as GerbilConfig } from "../types-LlyYILII.mjs";
+import "../outetts-DmR0NfSG.mjs";
+import { d as GerbilConfig } from "../types-Cpf-aOxX.mjs";
 import { Request, Response, Router } from "express";
 
 //#region src/frameworks/express.d.ts

package/dist/frameworks/express.d.mts.map

@@ -1 +1 @@
-{"version":3,"file":"express.d.mts","names":[],"sources":["../../src/frameworks/express.ts"],"sourcesContent":[],"mappings":";;;;;UAuBiB,oBAAA,SAA6B,KAAK;;;;;;;iBAQnC,MAAA,WAAgB,6BAgBE;;;;iBAqGlB,OAAA,WAAiB,6BAKZ,cAAc,aAAQ"}
+{"version":3,"file":"express.d.mts","names":[],"sources":["../../src/frameworks/express.ts"],"sourcesContent":[],"mappings":";;;;;;UAuBiB,oBAAA,SAA6B,KAAK;;;;;;;iBAQnC,MAAA,WAAgB,6BAgBE;;;;iBAqGlB,OAAA,WAAiB,6BAKZ,cAAc,aAAQ"}

package/dist/frameworks/express.mjs

@@ -1,5 +1,5 @@
 import { t as __require } from "../chunk-B9cbKln6.mjs";
-import { r as DEFAULT_MODEL, t as Gerbil } from "../gerbil-BMZ5wnM_.mjs";
+import { r as DEFAULT_MODEL, t as Gerbil } from "../gerbil-Du-1M2Wq.mjs";
 import "../utils-DKO55ZmZ.mjs";
 
 //#region src/frameworks/express.ts

package/dist/frameworks/fastify.d.mts

@@ -1,4 +1,5 @@
-import { d as GerbilConfig } from "../types-LlyYILII.mjs";
+import "../outetts-DmR0NfSG.mjs";
+import { d as GerbilConfig } from "../types-Cpf-aOxX.mjs";
 
 //#region src/frameworks/fastify.d.ts
 

package/dist/frameworks/fastify.d.mts.map

@@ -1 +1 @@
-{"version":3,"file":"fastify.d.mts","names":[],"sources":["../../src/frameworks/fastify.ts"],"sourcesContent":[],"mappings":";;;;UAiBiB,oBAAA,SAA6B;;;iBAIxB,YAAA,yBAAoC,uBAAyB"}
+{"version":3,"file":"fastify.d.mts","names":[],"sources":["../../src/frameworks/fastify.ts"],"sourcesContent":[],"mappings":";;;;;UAiBiB,oBAAA,SAA6B;;;iBAIxB,YAAA,yBAAoC,uBAAyB"}

package/dist/frameworks/fastify.mjs

@@ -1,4 +1,4 @@
-import { r as DEFAULT_MODEL, t as Gerbil } from "../gerbil-BMZ5wnM_.mjs";
+import { r as DEFAULT_MODEL, t as Gerbil } from "../gerbil-Du-1M2Wq.mjs";
 import "../utils-DKO55ZmZ.mjs";
 
 //#region src/frameworks/fastify.ts

package/dist/frameworks/hono.d.mts

@@ -1,4 +1,5 @@
-import { d as GerbilConfig } from "../types-LlyYILII.mjs";
+import "../outetts-DmR0NfSG.mjs";
+import { d as GerbilConfig } from "../types-Cpf-aOxX.mjs";
 import * as hono0 from "hono";
 import * as hono_types0 from "hono/types";
 

package/dist/frameworks/hono.d.mts.map

@@ -1 +1 @@
-{"version":3,"file":"hono.d.mts","names":[],"sources":["../../src/frameworks/hono.ts"],"sourcesContent":[],"mappings":";;;;;;UAiBiB,iBAAA,SAA0B;;;;iBAK3B,MAAA,WAAgB,0BAAsB,QAAA,KAAA,CAAA,KAAA,WAAA,CAAA,QAAA,EAAA,WAAA,CAAA,WAAA"}
+{"version":3,"file":"hono.d.mts","names":[],"sources":["../../src/frameworks/hono.ts"],"sourcesContent":[],"mappings":";;;;;;;AAsBsD,UALrC,iBAAA,SAA0B,YAKW,CAAA;;;;iBAAtC,MAAA,WAAgB,0BAAsB,QAAA,KAAA,CAAA,KAAA,WAAA,CAAA,QAAA,EAAA,WAAA,CAAA,WAAA"}

package/dist/frameworks/hono.mjs

@@ -1,4 +1,4 @@
-import { r as DEFAULT_MODEL, t as Gerbil } from "../gerbil-BMZ5wnM_.mjs";
+import { r as DEFAULT_MODEL, t as Gerbil } from "../gerbil-Du-1M2Wq.mjs";
 import "../utils-DKO55ZmZ.mjs";
 
 //#region src/frameworks/hono.ts

package/dist/frameworks/next.d.mts

@@ -1,6 +1,7 @@
-import { d as GerbilConfig } from "../types-LlyYILII.mjs";
-import "../index-HVS7yeLx.mjs";
-import { t as Gerbil } from "../gerbil-DkilSf1Q.mjs";
+import "../outetts-DmR0NfSG.mjs";
+import { d as GerbilConfig } from "../types-Cpf-aOxX.mjs";
+import "../index-C4kJrqs1.mjs";
+import { t as Gerbil } from "../gerbil-4Xid85_Z.mjs";
 
 //#region src/frameworks/next.d.ts
 

package/dist/frameworks/next.d.mts.map

@@ -1 +1 @@
-{"version":3,"file":"next.d.mts","names":[],"sources":["../../src/frameworks/next.ts"],"sourcesContent":[],"mappings":";;;;;;AAgD4C,UA9B3B,iBAAA,SAA0B,YA8BC,CAAA;EAAO;EA2CnC,OAAA,CAAA,EAAA,OAAc;;iBAhEf,WAAA,CAqBmB,OAAA,EArBE,iBAqBF,CAAA,EArBsB,OAqBtB,CArB8B,MAqB9B,CAAA;;;;AA+CI,iBAhDtB,OAAA,CAgDsB,OAAA,CAAA,EAhDL,iBAgDK,CAAA,EAAA,CAAA,GAAA,EA/CJ,OA+CI,EAAA,GA/CM,OA+CN,CA/Cc,QA+Cd,CAAA;;;;AAgBP,iBApBf,cAAA,CAoBe,OAAA,CAAA,EApBS,iBAoBT,CAAA,EAAA;EAiBV,QAAA,EAAA,CAAA,GAAA,EAhFa,OAgFb,EAAA,GAhFuB,OAgFvB,CAhF+B,QAgF/B,CAAA;EAAkB,IAAA,CAAA,GAAA,EAjCnB,OAiCmB,CAAA,EAjCT,OAiCS,CAjCD,QAiCC,CAAA;EAAR,KAAA,CAAA,GAAA,EAjBV,OAiBU,CAAA,EAjBA,OAiBA,CAjBQ,QAiBR,CAAA;EAAO,IAAA,CAAA,IAAA,EAAjB,OAAiB,CAAA,EAAP,OAAO,CAAC,QAAD,CAAA;AAWtC,CAAA;cAAa"}
+{"version":3,"file":"next.d.mts","names":[],"sources":["../../src/frameworks/next.ts"],"sourcesContent":[],"mappings":";;;;;;;AAgDoD,UA9BnC,iBAAA,SAA0B,YA8BS,CAAA;EAAR;EAAO,OAAA,CAAA,EAAA,OAAA;AA2CnD;iBAhEe,WAAA,CAgEyB,OAAA,EAhEJ,iBAgEI,CAAA,EAhEgB,OAgEhB,CAhEwB,MAgExB,CAAA;;;;AAIpB,iBAhDJ,OAAA,CAgDI,OAAA,CAAA,EAhDa,iBAgDb,CAAA,EAAA,CAAA,GAAA,EA/Cc,OA+Cd,EAAA,GA/CwB,OA+CxB,CA/CgC,QA+ChC,CAAA;;;;AAgBmB,iBApBvB,cAAA,CAoBuB,OAAA,CAAA,EApBC,iBAoBD,CAAA,EAAA;EAAR,QAAA,EAAA,CAAA,GAAA,EA/DG,OA+DH,EAAA,GA/Da,OA+Db,CA/DqB,QA+DrB,CAAA;EAiBV,IAAA,CAAA,GAAA,EAjCD,OAiCC,CAAA,EAjCS,OAiCT,CAjCiB,QAiCjB,CAAA;EAAkB,KAAA,CAAA,GAAA,EAjBlB,OAiBkB,CAAA,EAjBR,OAiBQ,CAjBA,QAiBA,CAAA;EAAR,IAAA,CAAA,IAAA,EAAV,OAAU,CAAA,EAAA,OAAA,CAAQ,QAAR,CAAA;CAAO;AAWzB,cAAA,MAIZ,EAAA"}

package/dist/frameworks/next.mjs

@@ -1,4 +1,4 @@
-import { r as DEFAULT_MODEL, t as Gerbil } from "../gerbil-BMZ5wnM_.mjs";
+import { r as DEFAULT_MODEL, t as Gerbil } from "../gerbil-Du-1M2Wq.mjs";
 import "../utils-DKO55ZmZ.mjs";
 
 //#region src/frameworks/next.ts

package/dist/frameworks/react.d.mts

@@ -1,4 +1,5 @@
-import { d as GerbilConfig, l as GenerateOptions, u as GenerateResult } from "../types-LlyYILII.mjs";
+import "../outetts-DmR0NfSG.mjs";
+import { d as GerbilConfig, l as GenerateOptions, u as GenerateResult } from "../types-Cpf-aOxX.mjs";
 
 //#region src/frameworks/react.d.ts
 

package/dist/frameworks/react.d.mts.map

@@ -1 +1 @@
-{"version":3,"file":"react.d.mts","names":[],"sources":["../../src/frameworks/react.ts"],"sourcesContent":[],"mappings":";;;;AA+N2B,UA1MV,gBAAA,SAAyB,YA0Mf,CAAA;EAAc;EAEhC,QAAA,CAAA,EAAA,MAAA;EAAK;EAIE,WAAO,CAAA,EAAA,OAAA;;KAxMX,eAAA;;uCAE2B,oBAAoB,QAAQ;;qCAG9B,oBAAoB;;;;SAMhD;;;;UAMC;;;;;;;iBASM,SAAA,WAAmB,mBAAwB;;;;KAuJ/C,OAAA;;;;UAKK,cAAA,SAAuB;;oBAEpB;;;;KAMR,aAAA;YACA;;;qBAGS,KAAA,CAAM,cAAc;;SAEhC;;;iBAIO,OAAA,WAAiB,iBAAsB"}
+{"version":3,"file":"react.d.mts","names":[],"sources":["../../src/frameworks/react.ts"],"sourcesContent":[],"mappings":";;;;;AA4NY,UAvMK,gBAAA,SAAyB,YAuM9B,CAAA;EAGS;EAAoB,QAAA,CAAA,EAAA,MAAA;EAEhC;EAAK,WAAA,CAAA,EAAA,OAAA;AAId;KAxMY,eAAA;;uCAE2B,oBAAoB,QAAQ;;qCAG9B,oBAAoB;;;;SAMhD;;;;UAMC;;;;;;;iBASM,SAAA,WAAmB,mBAAwB;;;;KAuJ/C,OAAA;;;;UAKK,cAAA,SAAuB;;oBAEpB;;;;KAMR,aAAA;YACA;;;qBAGS,KAAA,CAAM,cAAc;;SAEhC;;;iBAIO,OAAA,WAAiB,iBAAsB"}

package/dist/frameworks/trpc.d.mts

@@ -1,4 +1,5 @@
-import { d as GerbilConfig } from "../types-LlyYILII.mjs";
+import "../outetts-DmR0NfSG.mjs";
+import { d as GerbilConfig } from "../types-Cpf-aOxX.mjs";
 
 //#region src/frameworks/trpc.d.ts
 

package/dist/frameworks/trpc.d.mts.map

@@ -1 +1 @@
-{"version":3,"file":"trpc.d.mts","names":[],"sources":["../../src/frameworks/trpc.ts"],"sourcesContent":[],"mappings":";;;;UAqBiB,iBAAA,SAA0B;;;;iBAK3B,YAAA,mBAA8B"}
+{"version":3,"file":"trpc.d.mts","names":[],"sources":["../../src/frameworks/trpc.ts"],"sourcesContent":[],"mappings":";;;;;UAqBiB,iBAAA,SAA0B;;;;iBAK3B,YAAA,mBAA8B"}

package/dist/frameworks/trpc.mjs

@@ -1,4 +1,4 @@
-import { r as DEFAULT_MODEL, t as Gerbil } from "../gerbil-BMZ5wnM_.mjs";
+import { r as DEFAULT_MODEL, t as Gerbil } from "../gerbil-Du-1M2Wq.mjs";
 import "../utils-DKO55ZmZ.mjs";
 import { z } from "zod";
 

package/dist/{gerbil-DkilSf1Q.d.mts → gerbil-4Xid85_Z.d.mts}

@@ -1,5 +1,6 @@
-import { A as SpeakResult, C as PreloadOptions, D as SessionStats, E as SearchResult, I as TranscribeOptions, L as TranscribeResult, M as StreamingTranscriptionSession, N as SystemInfo, O as SimilarityResult, T as STTModelConfig, _ as LoadSTTOptions, a as EmbedResult, d as GerbilConfig, g as LoadOptions, h as JsonOptions, i as EmbedOptions, j as StreamingTranscriptionOptions, k as SpeakOptions, l as GenerateOptions, t as AudioChunk, u as GenerateResult, v as LoadTTSOptions, y as ModelConfig, z as VoiceInfo } from "./types-LlyYILII.mjs";
-import { a as GenerateObjectOptions, o as GenerateObjectResult } from "./index-HVS7yeLx.mjs";
+import { i as OuteSpeaker } from "./outetts-DmR0NfSG.mjs";
+import { A as SpeakResult, C as PreloadOptions, D as SessionStats, E as SearchResult, I as TranscribeOptions, L as TranscribeResult, M as StreamingTranscriptionSession, N as SystemInfo, O as SimilarityResult, T as STTModelConfig, _ as LoadSTTOptions, a as EmbedResult, d as GerbilConfig, g as LoadOptions, h as JsonOptions, i as EmbedOptions, j as StreamingTranscriptionOptions, k as SpeakOptions, l as GenerateOptions, t as AudioChunk, u as GenerateResult, v as LoadTTSOptions, y as ModelConfig, z as VoiceInfo } from "./types-Cpf-aOxX.mjs";
+import { a as GenerateObjectOptions, o as GenerateObjectResult } from "./index-C4kJrqs1.mjs";
 
 //#region src/core/gerbil.d.ts
 
@@ -14,6 +15,7 @@ declare class Gerbil {
   private nativeEmbedRepo;
   private nativeSTT;
   private nativeTTSEngine;
+  private nativeTTSRepo;
   private isVisionModel;
   private readonly queue;
   private readonly telemetry;
@@ -315,7 +317,7 @@ declare class Gerbil {
    * // result.audio = Float32Array PCM, result.sampleRate = 22050
    * ```
    */
-  loadTTS(_options?: LoadTTSOptions & {
+  loadTTS(options?: LoadTTSOptions & {
     model?: string;
   }): Promise<void>;
   /**
@@ -332,8 +334,29 @@ declare class Gerbil {
    * ```
    */
   speak(text: string, options?: SpeakOptions): Promise<SpeakResult>;
-  /** Lazily build (or reuse) the native Kani-TTS WebGPUEngine (default repo). */
+  /**
+   * Lazily build (or reuse) the native TTS WebGPUEngine for the requested model.
+   * Defaults to the small/fast Kani-TTS; selecting a different TTS repo (e.g. the
+   * OuteTTS many-voices engine) rebuilds the engine for that checkpoint.
+   */
   private ensureNativeTTSEngine;
+  /** List the voices the active (or default) TTS engine supports. */
+  ttsVoices(model?: string): Promise<string[]>;
+  /**
+   * OuteTTS voice cloning: encode a raw reference clip (mono PCM + transcript) into a
+   * reusable speaker, then synthesize new text in that voice via
+   * `speak(text, { model: DEFAULT_MODELS.ttsOute, cloneSpeaker })`.
+   *
+   * @example
+   * const g = createGerbil();
+   * const speaker = await g.encodeSpeaker({ audio, sampleRate, transcript }, DEFAULT_MODELS.ttsOute);
+   * const out = await g.speak("New words in the cloned voice.", { model: DEFAULT_MODELS.ttsOute, cloneSpeaker: speaker });
+   */
+  encodeSpeaker(ref: {
+    audio: Float32Array;
+    sampleRate: number;
+    transcript: string;
+  }, model?: string): Promise<OuteSpeaker>;
   /**
    * Stream speech generation. The native engine synthesizes the full clip, so a
    * single final audio chunk is yielded.
@@ -356,7 +379,9 @@ declare class Gerbil {
     device?: "webgpu" | "cpu";
   } | null;
   /**
-   * List available TTS models (native Kani-TTS-2).
+   * List available native TTS models: the default small/fast Kani-TTS and the
+   * selectable many-voices OuteTTS. Pass an OuteTTS `id` as `speak({ model })` /
+   * `loadTTS({ model })` to use it.
    */
   listTTSModels(): Promise<Array<{
     id: string;
@@ -485,4 +510,4 @@ declare class Gerbil {
 }
 //#endregion
 export { Gerbil as t };
-//# sourceMappingURL=gerbil-DkilSf1Q.d.mts.map
+//# sourceMappingURL=gerbil-4Xid85_Z.d.mts.map

package/dist/gerbil-4Xid85_Z.d.mts.map

@@ -0,0 +1 @@
+{"version":3,"file":"gerbil-4Xid85_Z.d.mts","names":[],"sources":["../src/core/gerbil.ts"],"sourcesContent":[],"mappings":";;;;;;AAgPsB,cAxBT,MAAA,CAwBS;EAgDC,QAAA,YAAA;EAIa,QAAA,WAAA;EA0BgB,iBAAA,MAAA;EAAmB,QAAA,KAAA;EAoLrD,QAAA,WAAA;EAyDsB,QAAA,YAAA;EAyCO,QAAA,iBAAA;EAAsB,QAAA,eAAA;EAqC7B,QAAA,SAAA;EAWM,QAAA,eAAA;EAAsB,QAAA,aAAA;EAgC5B,QAAA,aAAA;EAWM,iBAAA,KAAA;EAAsB,iBAAA,SAAA;EAgCvB,WAAA,CAAA,MAAA,CAAA,EA/dvB,YA+duB;EAWO,QAAA,WAAA;EAAsB;;;;;EAmQ7D,QAAA,YAAA;EACe,OAAA,UAAA,CAAA,CAAA,EA9rBL,WA8rBK,EAAA;EAAvB,OAAA,QAAA,CAAA,OAAA,EAAA,MAAA,CAAA,EA1rB+B,WA0rB/B,GAAA,SAAA;EA6MgD;;;;;;;;;;;;;;;;;;EA2M5B,SAAA,CAAA,OAAA,CAAA,EAAA,MAAA,EAAA,OAAA,CAAA,EAxjC2B,WAwjC3B,CAAA,EAxjC8C,OAwjC9C,CAAA,IAAA,CAAA;EACpB;;;;;;EAqHoB,QAAA,eAAA;EAA2C;;;EAoB/B,QAAA,CAAA,CAAA,EAAA,OAAA;EAA4B;;;EAoE/C,cAAA,CAAA,CAAA,EAAA,OAAA;EAEL;;;EAaO,YAAA,CAAA,CAAA,EAjmCF,WAimCE,GAAA,IAAA;EAAY;;;EA8C5B,aAAA,CAAA,CAAA,EAAA,QAAA,GAAA,KAAA,GAAA,MAAA;EADqB;;;;;EAsEd,QAAA,CAAA,CAAA,EAAA,MAAA;EAAe;;;EAErB,qBAAA,CAAA,CAAA,EAAA;IAoES,IAAA,EAAA,MAAA;IACD,MAAA,EAAA,MAAA;IAAR,IAAA,EAAA,MAAA;IAW4B,OAAA,EAAA,MAAA;EAAR,CAAA;EAyDZ;;;EAqDyB,kBAAA,CAAA,CAAA,EAAA,IAAA;EAmDX;;;;;;;;;;;;kCA94Ca;;;;;;;;;;;;;;;;;;;;;;;;0CAyCO,iBAAsB;;;;;kCAqC7B;;;;;;;yCAWM,iBAAsB;;;;;kCAgC5B;;;;;;;yCAWM,iBAAsB;;;;;uCAgCvB;;;;;;;+CAWO,iBAAsB;;;;;;gBAmCpD;;;;;;;;;;;;;;;qCAsBoB,kBAAuB,QAAQ;;;;;;;;;;;mCA0M5D,kBACR,uBAAuB;;;;;;;;;;;;;;;;;;;mCA6Ma,YAAY,KAAK,QAAQ;;;;;;;;;;;;;;;;;wDAgDrD,wBACR,QAAQ,qBAAqB;;;;gCAiBG,eAAoB,QAAQ;;;;;;;;;;wCAoDpB,eAAoB,QAAQ;;;;;;;;;;;;;;;;;;;;qDAkD5D,eACR,QADyB,gBAAA;;;;;;;;;;;;;;oDAmCjB;;MACR,QADoB,YAAA;;;;;;;;;;;mEAgCZ;;MACR,QADoB,YAAA;;;;cAyBX;;;;aAOD;;;;;;;;;;;;;;;;oBAsDY;;MAA2C;;;;6BAOjC,iBAAiB;;;;;;;;;;gCAaf,eAAoB,QAAQ;;;;;;;;6BAoD9B;;;;;;;;;;;;WAgBjB;;;sBAEb,QAAQ;;;;;sCAYA,eACR,eAAe,YAAY;;;;gBAchB;;;;;;;;;;;;;;;;;;mBA+BS,QACrB;;;;;;;;;;;;;;;;wCAiCyC,iBAAsB;;;;gDAQN,iBAAiB;;;;;;;;;;;;;;;;;;;;;;;;oBA4BnE,eAAe,sBACb,oBACR,QAAQ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;0CAoEC,gCACT,QAAQ;;;;mBAWY,QAAQ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;MAyD5B,QAAQ;;;;2BAmCoB;;;;;kCAkBK;;;;qBAmDX"}

package/dist/gerbil-BikD0xFU.mjs

@@ -0,0 +1,4 @@
+import { t as Gerbil } from "./gerbil-Du-1M2Wq.mjs";
+import "./utils-DKO55ZmZ.mjs";
+
+export { Gerbil };

package/dist/{gerbil-BMZ5wnM_.mjs → gerbil-Du-1M2Wq.mjs}

@@ -464,6 +464,7 @@ var Gerbil = class {
 	nativeEmbedRepo = null;
 	nativeSTT = null;
 	nativeTTSEngine = null;
+	nativeTTSRepo = null;
 	isVisionModel = false;
 	queue;
 	telemetry;
@@ -1340,8 +1341,8 @@ The JSON must conform to this schema: ${JSON.stringify(zodToJsonSchema(schema))}
 	* // result.audio = Float32Array PCM, result.sampleRate = 22050
 	* ```
 	*/
-	async loadTTS(_options = {}) {
-		await this.ensureNativeTTSEngine();
+	async loadTTS(options = {}) {
+		await this.ensureNativeTTSEngine(options.model);
 	}
 	/**
 	* Ensure TTS model is loaded (lazy loading)
@@ -1360,9 +1361,15 @@ The JSON must conform to this schema: ${JSON.stringify(zodToJsonSchema(schema))}
 	*/
 	async speak(text, options = {}) {
 		if (!this.preferNative()) throw new Error("Speech synthesis requires WebGPU. CPU/WASM and the legacy ONNX backend have been removed.");
-		const native = await this.ensureNativeTTSEngine();
+		const native = await this.ensureNativeTTSEngine(options.model);
 		const startTime = performance.now();
-		const out = await native.speak(text, {});
+		const out = await native.speak(text, {
+			voice: options.voice,
+			temperature: options.temperature,
+			topP: options.topP,
+			topK: options.topK,
+			cloneSpeaker: options.cloneSpeaker
+		});
 		return {
 			audio: out.pcm,
 			sampleRate: out.sampleRate,
@@ -1371,14 +1378,38 @@ The JSON must conform to this schema: ${JSON.stringify(zodToJsonSchema(schema))}
 			totalTime: performance.now() - startTime
 		};
 	}
-	/** Lazily build (or reuse) the native Kani-TTS WebGPUEngine (default repo). */
-	async ensureNativeTTSEngine() {
-		if (!this.nativeTTSEngine) {
-			const { WebGPUEngine, DEFAULT_MODELS } = await import("./gpu/index.mjs");
-			this.nativeTTSEngine = await WebGPUEngine.create({ repo: DEFAULT_MODELS.tts });
-		}
+	/**
+	* Lazily build (or reuse) the native TTS WebGPUEngine for the requested model.
+	* Defaults to the small/fast Kani-TTS; selecting a different TTS repo (e.g. the
+	* OuteTTS many-voices engine) rebuilds the engine for that checkpoint.
+	*/
+	async ensureNativeTTSEngine(model) {
+		const { WebGPUEngine, DEFAULT_MODELS } = await import("./gpu/index.mjs");
+		const repo = model ?? DEFAULT_MODELS.tts;
+		if (this.nativeTTSEngine && this.nativeTTSRepo === repo) return this.nativeTTSEngine;
+		if (this.nativeTTSEngine) this.nativeTTSEngine.destroy?.();
+		this.nativeTTSEngine = await WebGPUEngine.create({ repo });
+		this.nativeTTSRepo = repo;
 		return this.nativeTTSEngine;
 	}
+	/** List the voices the active (or default) TTS engine supports. */
+	async ttsVoices(model) {
+		return (await this.ensureNativeTTSEngine(model)).ttsVoices?.() ?? [];
+	}
+	/**
+	* OuteTTS voice cloning: encode a raw reference clip (mono PCM + transcript) into a
+	* reusable speaker, then synthesize new text in that voice via
+	* `speak(text, { model: DEFAULT_MODELS.ttsOute, cloneSpeaker })`.
+	*
+	* @example
+	* const g = createGerbil();
+	* const speaker = await g.encodeSpeaker({ audio, sampleRate, transcript }, DEFAULT_MODELS.ttsOute);
+	* const out = await g.speak("New words in the cloned voice.", { model: DEFAULT_MODELS.ttsOute, cloneSpeaker: speaker });
+	*/
+	async encodeSpeaker(ref, model) {
+		const { DEFAULT_MODELS } = await import("./gpu/index.mjs");
+		return (await this.ensureNativeTTSEngine(model ?? DEFAULT_MODELS.ttsOute)).encodeSpeaker(ref);
+	}
 	/**
 	* Stream speech generation. The native engine synthesizes the full clip, so a
 	* single final audio chunk is yielded.
@@ -1417,14 +1448,21 @@ The JSON must conform to this schema: ${JSON.stringify(zodToJsonSchema(schema))}
 		};
 	}
 	/**
-	* List available TTS models (native Kani-TTS-2).
+	* List available native TTS models: the default small/fast Kani-TTS and the
+	* selectable many-voices OuteTTS. Pass an OuteTTS `id` as `speak({ model })` /
+	* `loadTTS({ model })` to use it.
 	*/
 	async listTTSModels() {
 		return [{
 			id: this.ttsModelId,
-			description: "Kani-TTS-2 native WebGPU TTS",
+			description: "Kani-TTS native WebGPU TTS (small/fast, default)",
 			sampleRate: 22050,
 			voiceCount: 1
+		}, {
+			id: "OuteAI/OuteTTS-1.0-0.6B",
+			description: "OuteTTS-1.0-0.6B native WebGPU TTS (many voices + cloning)",
+			sampleRate: 24e3,
+			voiceCount: 6
 		}];
 	}
 	/**
@@ -1486,7 +1524,7 @@ The JSON must conform to this schema: ${JSON.stringify(zodToJsonSchema(schema))}
 	/** Lazily build (or reuse) the native MoonshineSTT engine (default repo). */
 	async ensureNativeSTT() {
 		if (!this.nativeSTT) {
-			const { MoonshineSTT } = await import("./moonshine-stt-Eei7fh3N.mjs");
+			const { MoonshineSTT } = await import("./moonshine-stt-fVw-UB4s.mjs");
 			this.nativeSTT = await MoonshineSTT.create();
 		}
 		return this.nativeSTT;
@@ -1653,4 +1691,4 @@ The JSON must conform to this schema: ${JSON.stringify(zodToJsonSchema(schema))}
 
 //#endregion
 export { resolveModel as a, configureGlobalCache as c, listBuiltinModels as i, getGlobalCache as l, BUILTIN_MODELS as n, ResponseCache as o, DEFAULT_MODEL as r, clearGlobalCache as s, Gerbil as t };
-//# sourceMappingURL=gerbil-BMZ5wnM_.mjs.map
+//# sourceMappingURL=gerbil-Du-1M2Wq.mjs.map