feat: Kokoro-js TTS support

This commit is contained in:
Timothy Jaeryang Baek
2025-02-09 23:42:27 -08:00
parent a22d1d5410
commit 205ce635f6
5 changed files with 388 additions and 78 deletions

View File

@@ -0,0 +1,70 @@
import WorkerInstance from '$lib/workers/kokoro.worker?worker';
export class KokoroWorker {
private worker: Worker | null = null;
private initialized: boolean = false;
private dtype: string;
constructor(dtype: string = 'fp32') {
this.dtype = dtype;
}
public async init() {
if (this.worker) {
console.warn('KokoroWorker is already initialized.');
return;
}
this.worker = new WorkerInstance();
return new Promise<void>((resolve, reject) => {
this.worker!.onmessage = (event) => {
const { status, error } = event.data;
if (status === 'init:complete') {
this.initialized = true;
resolve();
} else if (status === 'init:error') {
console.error(error);
this.initialized = false;
reject(new Error(error));
}
};
this.worker!.postMessage({
type: 'init',
payload: { dtype: this.dtype }
});
});
}
public async generate({ text, voice }: { text: string; voice: string }): Promise<string> {
if (!this.initialized || !this.worker) {
throw new Error('KokoroTTS Worker is not initialized yet.');
}
return new Promise<string>((resolve, reject) => {
this.worker.postMessage({ type: 'generate', payload: { text, voice } });
const handleMessage = (event: MessageEvent) => {
if (event.data.status === 'generate:complete') {
this.worker!.removeEventListener('message', handleMessage);
resolve(event.data.audioUrl);
} else if (event.data.status === 'generate:error') {
this.worker!.removeEventListener('message', handleMessage);
reject(new Error(event.data.error));
}
};
this.worker.addEventListener('message', handleMessage);
});
}
public terminate() {
if (this.worker) {
this.worker.terminate();
this.worker = null;
this.initialized = false;
}
}
}

View File

@@ -0,0 +1,53 @@
import { KokoroTTS } from 'kokoro-js';
let tts;
let isInitialized = false; // Flag to track initialization status
const DEFAULT_MODEL_ID = 'onnx-community/Kokoro-82M-v1.0-ONNX'; // Default model
self.onmessage = async (event) => {
const { type, payload } = event.data;
if (type === 'init') {
let { model_id, dtype } = payload;
model_id = model_id || DEFAULT_MODEL_ID; // Use default model if none provided
self.postMessage({ status: 'init:start' });
try {
tts = await KokoroTTS.from_pretrained(model_id, {
dtype,
device: !!navigator?.gpu ? 'webgpu' : 'wasm' // Detect WebGPU
});
isInitialized = true; // Mark as initialized after successful loading
self.postMessage({ status: 'init:complete' });
} catch (error) {
isInitialized = false; // Ensure it's marked as false on failure
self.postMessage({ status: 'init:error', error: error.message });
}
}
if (type === 'generate') {
if (!isInitialized || !tts) {
// Ensure model is initialized
self.postMessage({ status: 'generate:error', error: 'TTS model not initialized' });
return;
}
const { text, voice } = payload;
self.postMessage({ status: 'generate:start' });
try {
const rawAudio = await tts.generate(text, { voice });
const blob = await rawAudio.toBlob();
const blobUrl = URL.createObjectURL(blob);
self.postMessage({ status: 'generate:complete', audioUrl: blobUrl });
} catch (error) {
self.postMessage({ status: 'generate:error', error: error.message });
}
}
if (type === 'status') {
// Respond with the current initialization status
self.postMessage({ status: 'status:check', initialized: isInitialized });
}
};