Fix Kokoro TTS integration issues: Remove API key requirement and ensure system-specific options display correctly
This commit is contained in:
@@ -0,0 +1,270 @@
|
||||
/**
|
||||
* ElevenLabsTTSModule
|
||||
* Provides TTS via ElevenLabs API
|
||||
*/
|
||||
import { ApiTTSModuleBase } from './api-tts-module-base.js';
|
||||
|
||||
export class ElevenLabsTTSModule extends ApiTTSModuleBase {
|
||||
constructor() {
|
||||
super('elevenlabs', 'ElevenLabs TTS');
|
||||
|
||||
// Voice options specific to ElevenLabs
|
||||
this.voiceOptions = {
|
||||
voice: 'pNInz6obpgDQGcFmaJgB', // Default voice ID for ElevenLabs
|
||||
model: 'eleven_multilingual_v2', // Use the multilingual model
|
||||
speed: 1.0
|
||||
};
|
||||
}
|
||||
|
||||
/**
|
||||
* Initialize the ElevenLabs TTS module
|
||||
* @returns {Promise<boolean>} - Resolves with success status
|
||||
*/
|
||||
async initialize() {
|
||||
try {
|
||||
this.reportProgress(10, 'Initializing ElevenLabs TTS');
|
||||
|
||||
// Initialize parent
|
||||
const parentInit = await super.initialize();
|
||||
if (!parentInit) {
|
||||
console.error('ElevenLabs TTS: Parent initialization failed');
|
||||
return false;
|
||||
}
|
||||
|
||||
// Get required dependencies
|
||||
const persistenceManager = this.getModule('persistence-manager');
|
||||
if (!persistenceManager) {
|
||||
console.error('ElevenLabs TTS: Required dependency persistence-manager not found');
|
||||
return false;
|
||||
}
|
||||
|
||||
// Check for API key
|
||||
const apiKey = persistenceManager.getPreference('elevenlabs', 'api_key', '');
|
||||
if (!apiKey) {
|
||||
console.error('ElevenLabs TTS: API key not configured');
|
||||
return false;
|
||||
}
|
||||
|
||||
// Load voices from ElevenLabs
|
||||
try {
|
||||
this.reportProgress(50, 'Loading ElevenLabs voices');
|
||||
await this.loadVoices(apiKey);
|
||||
} catch (error) {
|
||||
console.error('ElevenLabs TTS: Failed to load voices:', error);
|
||||
return false;
|
||||
}
|
||||
|
||||
// Load preferences
|
||||
const preferredVoice = persistenceManager.getPreference('elevenlabs', 'voice', this.voiceOptions.voice);
|
||||
if (preferredVoice) {
|
||||
this.voiceOptions.voice = preferredVoice;
|
||||
}
|
||||
|
||||
const preferredModel = persistenceManager.getPreference('elevenlabs', 'model', this.voiceOptions.model);
|
||||
if (preferredModel) {
|
||||
this.voiceOptions.model = preferredModel;
|
||||
}
|
||||
|
||||
const preferredSpeed = persistenceManager.getPreference('elevenlabs', 'speed', this.voiceOptions.speed);
|
||||
if (typeof preferredSpeed === 'number') {
|
||||
this.voiceOptions.speed = preferredSpeed;
|
||||
}
|
||||
|
||||
this.isReady = true;
|
||||
this.reportProgress(100, 'ElevenLabs TTS initialized');
|
||||
return true;
|
||||
} catch (error) {
|
||||
console.error('ElevenLabs TTS: Initialization error:', error);
|
||||
this.isReady = false;
|
||||
return false;
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Get the default API base URL for ElevenLabs
|
||||
* @returns {string} - Default API base URL
|
||||
*/
|
||||
getDefaultApiBaseUrl() {
|
||||
return 'https://api.elevenlabs.io/v1';
|
||||
}
|
||||
|
||||
/**
|
||||
* Load available voices from ElevenLabs API
|
||||
* @param {string} apiKey - API key for authentication
|
||||
* @returns {Promise<boolean>} - Resolves with success status
|
||||
*/
|
||||
async loadVoices(apiKey) {
|
||||
// Set default voices that will be used if API call fails
|
||||
this.voices = [
|
||||
{ id: 'pNInz6obpgDQGcFmaJgB', name: 'Rachel', language: 'en' },
|
||||
{ id: '21m00Tcm4TlvDq8ikWAM', name: 'Adam', language: 'en' },
|
||||
{ id: 'AZnzlk1XvdvUeBnXmlld', name: 'Antoni', language: 'en' },
|
||||
{ id: 'EXAVITQu4vr4xnSDxMaL', name: 'Bella', language: 'en' },
|
||||
{ id: 'ErXwobaYiN019PkySvjV', name: 'Daniel', language: 'en' }
|
||||
];
|
||||
|
||||
// Only load from API if we have an API key
|
||||
if (!apiKey) {
|
||||
return true;
|
||||
}
|
||||
|
||||
try {
|
||||
const response = await fetch(`${this.apiBaseUrl}/voices`, {
|
||||
method: 'GET',
|
||||
headers: {
|
||||
'xi-api-key': apiKey,
|
||||
'Content-Type': 'application/json'
|
||||
}
|
||||
});
|
||||
|
||||
if (!response.ok) {
|
||||
console.error(`ElevenLabs TTS: API error: ${response.status} ${response.statusText}`);
|
||||
return true; // Use defaults, but don't fail initialization
|
||||
}
|
||||
|
||||
const data = await response.json();
|
||||
|
||||
if (data && data.voices && Array.isArray(data.voices)) {
|
||||
// Transform API response to our internal format
|
||||
this.voices = data.voices.map(voice => ({
|
||||
id: voice.voice_id,
|
||||
name: voice.name,
|
||||
language: 'en', // ElevenLabs doesn't provide language info
|
||||
preview: voice.preview_url
|
||||
}));
|
||||
|
||||
return true;
|
||||
}
|
||||
} catch (error) {
|
||||
console.error('ElevenLabs TTS: Error loading voices:', error);
|
||||
}
|
||||
|
||||
// If API call failed, we still return true since we have default voices
|
||||
return true;
|
||||
}
|
||||
|
||||
/**
|
||||
* Select a voice for the given locale
|
||||
* @param {string} locale - Locale code
|
||||
* @returns {boolean} - Success status
|
||||
*/
|
||||
selectVoiceForLocale(locale) {
|
||||
if (!this.voices || this.voices.length === 0) {
|
||||
return this.selectDefaultVoice();
|
||||
}
|
||||
|
||||
// ElevenLabs doesn't provide language info for voices
|
||||
// Simply use the first voice as default
|
||||
return this.selectDefaultVoice();
|
||||
}
|
||||
|
||||
/**
|
||||
* Generate speech audio data using ElevenLabs API
|
||||
* @param {string} text - Text to generate speech for
|
||||
* @returns {Promise<Object>} - Audio data object
|
||||
*/
|
||||
async generateSpeechAudio(text) {
|
||||
// Don't attempt to call the API if no API key is set or text is empty
|
||||
if (!text || !this.apiKey) {
|
||||
return { success: false, reason: 'missing_api_key_or_text' };
|
||||
}
|
||||
|
||||
try {
|
||||
// Process the text
|
||||
const processedText = this.preprocessText(text);
|
||||
|
||||
// Create request payload
|
||||
const payload = {
|
||||
text: processedText,
|
||||
model_id: this.voiceOptions.model || 'eleven_multilingual_v2',
|
||||
voice_settings: {
|
||||
stability: 0.5,
|
||||
similarity_boost: 0.75,
|
||||
style: 0.0,
|
||||
use_speaker_boost: true,
|
||||
speed: this.voiceOptions.speed || 1.0
|
||||
}
|
||||
};
|
||||
|
||||
// Make API request
|
||||
const response = await fetch(`${this.apiBaseUrl}/text-to-speech/${this.voiceOptions.voice}?optimize_streaming_latency=0`, {
|
||||
method: 'POST',
|
||||
headers: {
|
||||
'Content-Type': 'application/json',
|
||||
'xi-api-key': this.apiKey,
|
||||
'Accept': 'audio/wav'
|
||||
},
|
||||
body: JSON.stringify(payload)
|
||||
});
|
||||
|
||||
if (!response.ok) {
|
||||
throw new Error(`API error: ${response.status} ${response.statusText}`);
|
||||
}
|
||||
|
||||
// Get audio blob from response
|
||||
const audioBlob = await response.blob();
|
||||
|
||||
// Convert to array buffer for consistency with other modules
|
||||
const arrayBuffer = await audioBlob.arrayBuffer();
|
||||
|
||||
return {
|
||||
success: true,
|
||||
audioData: arrayBuffer
|
||||
};
|
||||
} catch (error) {
|
||||
console.error('ElevenLabs TTS: Error generating speech:', error);
|
||||
return {
|
||||
success: false,
|
||||
reason: 'api_error',
|
||||
error: error.message
|
||||
};
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Set voice options
|
||||
* @param {Object} options - Voice options
|
||||
*/
|
||||
setVoiceOptions(options = {}) {
|
||||
// Call parent method for common options
|
||||
if (options.voice) {
|
||||
this.voiceOptions.voice = options.voice;
|
||||
|
||||
// Save voice preference
|
||||
const persistenceManager = this.getModule('persistence-manager');
|
||||
if (persistenceManager) {
|
||||
persistenceManager.updatePreference('tts', 'elevenlabs_voice', options.voice);
|
||||
}
|
||||
}
|
||||
|
||||
if (typeof options.speed === 'number') {
|
||||
this.voiceOptions.speed = Math.max(0.5, Math.min(2.0, options.speed));
|
||||
}
|
||||
|
||||
// Handle ElevenLabs-specific options
|
||||
if (options.model) {
|
||||
this.voiceOptions.model = options.model;
|
||||
|
||||
// Save model preference
|
||||
const persistenceManager = this.getModule('persistence-manager');
|
||||
if (persistenceManager) {
|
||||
persistenceManager.updatePreference('tts', 'elevenlabs_model', options.model);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Register the module with the module registry
|
||||
// Module registry MUST be accessed via window, not direct import
|
||||
if (window.moduleRegistry) {
|
||||
try {
|
||||
// Create instance first, then register it
|
||||
const elevenLabsTTSModule = new ElevenLabsTTSModule();
|
||||
window.moduleRegistry.register(elevenLabsTTSModule);
|
||||
console.log('ElevenLabs TTS Module registered successfully');
|
||||
} catch (err) {
|
||||
console.error('Failed to register ElevenLabs TTS Module:', err);
|
||||
}
|
||||
} else {
|
||||
console.error('Module registry not available when attempting to register ElevenLabs TTS Module');
|
||||
}
|
||||
Reference in New Issue
Block a user