Kokoro: chunk generation, add pre-process func #3412

1b817cd89747f2d1ffda68dfe63d36c25f777395

Cohee <18619528+Cohee1207@users.noreply.github.com>

public/scripts/extensions/tts/kokoro.js+32 -6 Viewed

 import { debounce_timeout } from '../../constants.js';
-import { debounceAsync } from '../../utils.js';
+import { debounceAsync, splitRecursive } from '../../utils.js';
 import { getPreviewString, saveTtsProviderSettings } from './index.js';
 export class KokoroTtsProvider {

         this.initTtsDebounced = debounceAsync(this.initializeWorker.bind(this), debounce_timeout.relaxed);
+    }
+    /**
+     * Perform any text processing before passing to TTS engine.
+     * @param {string} text Input text
+     * @returns {string} Processed text
+     */
+    processText(text) {
+        // TILDE!
+        text = text.replace(/~/g, '.');
+        return text;
+    }
     async loadSettings(settings) {
         if (settings.modelId !== undefined) this.settings.modelId = settings.modelId;
         if (settings.dtype !== undefined) this.settings.dtype = settings.dtype;

         const voice = this.getVoice(voiceId);
         const previewText = getPreviewString(voice.lang);
-        const response = await this.generateTts(previewText, voiceId);
+        for await (const response of this.generateTts(previewText, voiceId)) {
             const audio = await response.blob();
             const url = URL.createObjectURL(audio);
+            await new Promise(resolve => {
                 const audioElement = new Audio();
                 audioElement.src = url;
                 audioElement.play();
-        audioElement.onended = () => URL.revokeObjectURL(url);
+                audioElement.onended = () => resolve();
+            });
+            URL.revokeObjectURL(url);
+        }
+    }
     getVoiceDisplayName(voiceId) {

         };
+    }
-    async generateTts(text, voiceId) {
+    /**
+     * Generate TTS audio for the given text using the specified voice.
+     * @param {string} text Text to generate
+     * @param {string} voiceId Voice ID
+     * @returns {AsyncGenerator<Response>} Audio response generator
+     */
+    async* generateTts(text, voiceId) {
         if (!this.ready || !this.worker) {
             console.log('TTS not ready, initializing...');
             await this.initializeWorker();

         const voice = this.getVoice(voiceId);
         const requestId = this.nextRequestId++;
-        return new Promise((resolve, reject) => {
+        const chunkSize = 400;
+        const chunks = splitRecursive(text, chunkSize, ['\n\n', '\n', '.', '?', '!', ',', ' ', '']);
+        for (const chunk of chunks) {
+            yield await new Promise((resolve, reject) => {
                 // Store the promise callbacks
                 this.pendingRequests.set(requestId, { resolve, reject });

                 this.worker.postMessage({
                     action: 'generateTts',
                     data: {
-                    text,
+                        text: chunk,
                         voice: voice.voice_id,
                         speakingRate: this.settings.speakingRate || 1.0,
                         requestId,

                 });
             });
+        }
+    }
     dispose() {
         // Clean up the worker when the provider is disposed

public/scripts/utils.js+4 -0 Viewed

     return result;
+}
+export function splitSentences(input, length) {
+    var pattRegex = new RegExp(`^[\\s\\S]{${Math.floor(length / 2)},${length}}[.!?,]{1}|^[\\s\\S]{1,${length}}$|^[\\s\\S]{1,${length}}`);
+}
 /**
  * Checks if a string is a valid data URL.
  * @param {string} str The string to check.