@tryhamster/gerbil 1.3.1 → 1.4.0

package/dist/{defaults-CudhglS1.mjs → defaults-KnrRSIeJ.mjs}

@@ -10,6 +10,7 @@ const DEFAULT_MODELS = {
 	embedding: "mlx-community/embeddinggemma-300m-4bit",
 	tts: "nineninesix/kani-tts-450m-0.2-ft",
 	ttsOute: "OuteAI/OuteTTS-1.0-0.6B",
+	ttsParler: "parler-tts/parler-tts-mini-v1",
 	stt: "UsefulSensors/moonshine-base"
 };
 /**
@@ -25,14 +26,26 @@ const OUTETTS_ASSETS = {
 	speakersBaseURL: `${OUTETTS_ASSET_BASE}/speakers`,
 	defaultVoice: "en-female-1-neutral"
 };
-/** Bundled OuteTTS preset voices (the edwko/OuteTTS v3 default_speakers set). */
+/**
+* Bundled OuteTTS preset voices.
+*
+* Provenance (honest labelling):
+*  - `en-female-1-neutral` is the ONE AUTHENTIC OuteAI voice — the OuteTTS v3 bundled
+*    default speaker (default_speakers/en-female-1-neutral), copied verbatim.
+*  - `gerbil-voice-2..6` are DERIVED voices: distinct timbres produced by running
+*    pitch/tempo-varied renderings of that authentic clip back through the REAL
+*    DAC.speech encode + prosody pipeline (scripts/engine/stage-outetts-assets.py).
+*    They are genuine DAC-encoded OuteSpeakers, but they are NOT OuteAI's own
+*    `en-male-*` voices, so they are not labelled as such. (Earlier builds mislabelled
+*    them as `en-male-2-neutral` etc.; renamed for honesty.)
+*/
 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"
+	"gerbil-voice-2",
+	"gerbil-voice-3",
+	"gerbil-voice-4",
+	"gerbil-voice-5",
+	"gerbil-voice-6"
 ];
 /** Resolve the model repo for a set of options, falling back to the defaults. */
 function resolveDefaultRepo(opts) {
@@ -44,4 +57,4 @@ function resolveDefaultRepo(opts) {
 
 //#endregion
 export { resolveDefaultRepo as i, OUTETTS_ASSETS as n, OUTETTS_PRESET_VOICES as r, DEFAULT_MODELS as t };
-//# sourceMappingURL=defaults-CudhglS1.mjs.map
+//# sourceMappingURL=defaults-KnrRSIeJ.mjs.map

package/dist/defaults-KnrRSIeJ.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"defaults-KnrRSIeJ.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  /**\n   * Selectable describe-your-voice TTS (Parler-TTS-mini-v1: Flan-T5 encoder +\n   * cross-attending decoder LM + dac_44khz). Pass this as the `repo`/`model` to\n   * engine.speak()/useTTS with a `describeVoice` natural-language description.\n   */\n  ttsParler: \"parler-tts/parler-tts-mini-v1\",\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/**\n * Bundled OuteTTS preset voices.\n *\n * Provenance (honest labelling):\n *  - `en-female-1-neutral` is the ONE AUTHENTIC OuteAI voice — the OuteTTS v3 bundled\n *    default speaker (default_speakers/en-female-1-neutral), copied verbatim.\n *  - `gerbil-voice-2..6` are DERIVED voices: distinct timbres produced by running\n *    pitch/tempo-varied renderings of that authentic clip back through the REAL\n *    DAC.speech encode + prosody pipeline (scripts/engine/stage-outetts-assets.py).\n *    They are genuine DAC-encoded OuteSpeakers, but they are NOT OuteAI's own\n *    `en-male-*` voices, so they are not labelled as such. (Earlier builds mislabelled\n *    them as `en-male-2-neutral` etc.; renamed for honesty.)\n */\nexport const OUTETTS_PRESET_VOICES = [\n  \"en-female-1-neutral\",\n  \"gerbil-voice-2\",\n  \"gerbil-voice-3\",\n  \"gerbil-voice-4\",\n  \"gerbil-voice-5\",\n  \"gerbil-voice-6\",\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;CAMT,WAAW;CAEX,KAAK;CACN;;;;;;;;AASD,MAAM,qBACJ;AACF,MAAa,iBAAiB;CAE5B,iBAAiB,GAAG,mBAAmB;CAEvC,iBAAiB,GAAG,mBAAmB;CAEvC,cAAc;CACf;;;;;;;;;;;;;;AAeD,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

@@ -1,5 +1,5 @@
-import "../outetts-DmR0NfSG.mjs";
-import { d as GerbilConfig } from "../types-Cpf-aOxX.mjs";
+import "../outetts-Cr5oGk4z.mjs";
+import { d as GerbilConfig } from "../types-CjG36Vzj.mjs";
 import { Request, Response, Router } from "express";
 
 //#region src/frameworks/express.d.ts

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

package/dist/frameworks/fastify.d.mts

@@ -1,5 +1,5 @@
-import "../outetts-DmR0NfSG.mjs";
-import { d as GerbilConfig } from "../types-Cpf-aOxX.mjs";
+import "../outetts-Cr5oGk4z.mjs";
+import { d as GerbilConfig } from "../types-CjG36Vzj.mjs";
 
 //#region src/frameworks/fastify.d.ts
 

package/dist/frameworks/fastify.mjs

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

package/dist/frameworks/hono.d.mts

@@ -1,5 +1,5 @@
-import "../outetts-DmR0NfSG.mjs";
-import { d as GerbilConfig } from "../types-Cpf-aOxX.mjs";
+import "../outetts-Cr5oGk4z.mjs";
+import { d as GerbilConfig } from "../types-CjG36Vzj.mjs";
 import * as hono0 from "hono";
 import * as hono_types0 from "hono/types";
 

package/dist/frameworks/hono.mjs

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

package/dist/frameworks/next.d.mts

@@ -1,7 +1,7 @@
-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";
+import "../outetts-Cr5oGk4z.mjs";
+import { d as GerbilConfig } from "../types-CjG36Vzj.mjs";
+import "../index-BdbeBDTz.mjs";
+import { t as Gerbil } from "../gerbil-_68_JSGV.mjs";
 
 //#region src/frameworks/next.d.ts
 

package/dist/frameworks/next.mjs

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

package/dist/frameworks/react.d.mts

@@ -1,5 +1,5 @@
-import "../outetts-DmR0NfSG.mjs";
-import { d as GerbilConfig, l as GenerateOptions, u as GenerateResult } from "../types-Cpf-aOxX.mjs";
+import "../outetts-Cr5oGk4z.mjs";
+import { d as GerbilConfig, l as GenerateOptions, u as GenerateResult } from "../types-CjG36Vzj.mjs";
 
 //#region src/frameworks/react.d.ts
 

package/dist/frameworks/trpc.d.mts

@@ -1,5 +1,5 @@
-import "../outetts-DmR0NfSG.mjs";
-import { d as GerbilConfig } from "../types-Cpf-aOxX.mjs";
+import "../outetts-Cr5oGk4z.mjs";
+import { d as GerbilConfig } from "../types-CjG36Vzj.mjs";
 
 //#region src/frameworks/trpc.d.ts
 

package/dist/frameworks/trpc.mjs

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

package/dist/gerbil-CUzVDAU7.mjs

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

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

@@ -1368,7 +1368,8 @@ The JSON must conform to this schema: ${JSON.stringify(zodToJsonSchema(schema))}
 			temperature: options.temperature,
 			topP: options.topP,
 			topK: options.topK,
-			cloneSpeaker: options.cloneSpeaker
+			cloneSpeaker: options.cloneSpeaker,
+			describeVoice: options.describeVoice
 		});
 		return {
 			audio: out.pcm,
@@ -1411,6 +1412,28 @@ The JSON must conform to this schema: ${JSON.stringify(zodToJsonSchema(schema))}
 		return (await this.ensureNativeTTSEngine(model ?? DEFAULT_MODELS.ttsOute)).encodeSpeaker(ref);
 	}
 	/**
+	* Invent a natural-language voice description for Parler-TTS. If a text model is
+	* loaded it writes a richer one-line description from `hint` (e.g. "an excited
+	* sports announcer"); otherwise it falls back to a host-side template bank. Pass the
+	* result as `speak(text, { model: DEFAULT_MODELS.ttsParler, describeVoice })`.
+	*
+	* @example
+	* const desc = await g.inventVoiceDescription("a calm meditation guide");
+	* await g.speak("Breathe in.", { model: DEFAULT_MODELS.ttsParler, describeVoice: desc });
+	*/
+	async inventVoiceDescription(hint) {
+		if (this.isLoaded()) try {
+			const prompt = "Write ONE concise sentence describing a speaking voice for a text-to-speech model. Mention gender, tone, pace, and recording quality. Reply with only the sentence." + (hint ? ` The voice should sound like: ${hint}.` : "");
+			const line = (await this.generate(prompt, {
+				maxTokens: 60,
+				temperature: .9
+			})).text.trim().split("\n")[0].trim();
+			if (line.length > 0) return line;
+		} catch {}
+		const { ParlerTTS } = await import("./gpu/index.mjs");
+		return ParlerTTS.randomVoiceDescription();
+	}
+	/**
 	* Stream speech generation. The native engine synthesizes the full clip, so a
 	* single final audio chunk is yielded.
 	*/
@@ -1524,7 +1547,7 @@ 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-fVw-UB4s.mjs");
+			const { MoonshineSTT } = await import("./moonshine-stt-QoNJeCzR.mjs");
 			this.nativeSTT = await MoonshineSTT.create();
 		}
 		return this.nativeSTT;
@@ -1691,4 +1714,4 @@ 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-Du-1M2Wq.mjs.map
+//# sourceMappingURL=gerbil-DGrCwGtJ.mjs.map