SillyTavern/public/scripts/extensions/tts/silerotts.js

177 lines
5.6 KiB
JavaScript
Raw Normal View History

2023-12-02 20:11:06 +01:00
import { doExtrasFetch, getApiUrl, modules } from '../../extensions.js';
import { saveTtsProviderSettings } from './index.js';
2023-07-20 19:32:15 +02:00
2023-12-02 20:11:06 +01:00
export { SileroTtsProvider };
2023-07-20 19:32:15 +02:00
class SileroTtsProvider {
//########//
// Config //
//########//
2023-12-02 20:11:06 +01:00
settings;
ready = false;
voices = [];
separator = ' .. ';
2023-07-20 19:32:15 +02:00
defaultSettings = {
2023-12-02 19:04:51 +01:00
provider_endpoint: 'http://localhost:8001/tts',
2023-12-02 21:06:57 +01:00
voiceMap: {},
2023-12-02 20:11:06 +01:00
};
2023-07-20 19:32:15 +02:00
get settingsHtml() {
let html = `
<label for="silero_tts_endpoint">Provider Endpoint:</label>
<input id="silero_tts_endpoint" type="text" class="text_pole" maxlength="250" value="${this.defaultSettings.provider_endpoint}"/>
<span>
<span>Use <a target="_blank" href="https://github.com/SillyTavern/SillyTavern-extras">SillyTavern Extras API</a> or <a target="_blank" href="https://github.com/ouoertheo/silero-api-server">Silero TTS Server</a>.</span>
2023-12-02 20:11:06 +01:00
`;
return html;
2023-07-20 19:32:15 +02:00
}
onSettingsChange() {
// Used when provider settings are updated from UI
2023-12-02 20:11:06 +01:00
this.settings.provider_endpoint = $('#silero_tts_endpoint').val();
saveTtsProviderSettings();
this.refreshSession();
2023-07-20 19:32:15 +02:00
}
async loadSettings(settings) {
2023-07-20 19:32:15 +02:00
// Pupulate Provider UI given input settings
if (Object.keys(settings).length == 0) {
2023-12-02 20:11:06 +01:00
console.info('Using default TTS Provider settings');
2023-07-20 19:32:15 +02:00
}
// Only accept keys defined in defaultSettings
2023-12-02 20:11:06 +01:00
this.settings = this.defaultSettings;
2023-07-20 19:32:15 +02:00
2023-11-28 18:24:26 +01:00
for (const key in settings) {
if (key in this.settings) {
2023-12-02 20:11:06 +01:00
this.settings[key] = settings[key];
2023-07-20 19:32:15 +02:00
} else {
2023-12-02 20:11:06 +01:00
throw `Invalid setting passed to TTS Provider: ${key}`;
2023-07-20 19:32:15 +02:00
}
}
const apiCheckInterval = setInterval(() => {
// Use Extras API if TTS support is enabled
if (modules.includes('tts') || modules.includes('silero-tts')) {
const baseUrl = new URL(getApiUrl());
baseUrl.pathname = '/api/tts';
this.settings.provider_endpoint = baseUrl.toString();
$('#silero_tts_endpoint').val(this.settings.provider_endpoint);
clearInterval(apiCheckInterval);
}
}, 2000);
2023-12-02 20:11:06 +01:00
$('#silero_tts_endpoint').val(this.settings.provider_endpoint);
$('#silero_tts_endpoint').on('input', () => { this.onSettingsChange(); });
this.refreshSession();
2023-08-22 15:30:33 +02:00
2023-12-02 20:11:06 +01:00
await this.checkReady();
2023-08-22 15:30:33 +02:00
2023-12-02 20:11:06 +01:00
console.debug('SileroTTS: Settings loaded');
2023-07-20 19:32:15 +02:00
}
2023-08-22 15:30:33 +02:00
// Perform a simple readiness check by trying to fetch voiceIds
2023-11-28 18:24:26 +01:00
async checkReady() {
2023-12-02 20:11:06 +01:00
await this.fetchTtsVoiceObjects();
2023-08-22 15:30:33 +02:00
}
2023-07-20 19:32:15 +02:00
2023-08-26 05:52:26 +02:00
async onRefreshClick() {
2023-12-02 20:11:06 +01:00
return;
2023-07-20 19:32:15 +02:00
}
2023-11-28 18:24:26 +01:00
async refreshSession() {
2023-12-02 20:11:06 +01:00
await this.initSession();
}
2023-07-20 19:32:15 +02:00
//#################//
// TTS Interfaces //
//#################//
async getVoice(voiceName) {
if (this.voices.length == 0) {
2023-12-02 20:11:06 +01:00
this.voices = await this.fetchTtsVoiceObjects();
2023-07-20 19:32:15 +02:00
}
const match = this.voices.filter(
2023-12-02 21:06:57 +01:00
sileroVoice => sileroVoice.name == voiceName,
2023-12-02 20:11:06 +01:00
)[0];
2023-07-20 19:32:15 +02:00
if (!match) {
2023-12-02 20:11:06 +01:00
throw `TTS Voice name ${voiceName} not found`;
2023-07-20 19:32:15 +02:00
}
2023-12-02 20:11:06 +01:00
return match;
2023-07-20 19:32:15 +02:00
}
2023-11-28 18:24:26 +01:00
async generateTts(text, voiceId) {
2023-12-02 20:11:06 +01:00
const response = await this.fetchTtsGeneration(text, voiceId);
return response;
2023-07-20 19:32:15 +02:00
}
//###########//
// API CALLS //
//###########//
2023-08-26 05:52:26 +02:00
async fetchTtsVoiceObjects() {
2023-12-02 20:11:06 +01:00
const response = await doExtrasFetch(`${this.settings.provider_endpoint}/speakers`);
2023-07-20 19:32:15 +02:00
if (!response.ok) {
2023-12-02 20:11:06 +01:00
throw new Error(`HTTP ${response.status}: ${await response.json()}`);
2023-07-20 19:32:15 +02:00
}
2023-12-02 20:11:06 +01:00
const responseJson = await response.json();
return responseJson;
2023-07-20 19:32:15 +02:00
}
async fetchTtsGeneration(inputText, voiceId) {
2023-12-02 20:11:06 +01:00
console.info(`Generating new TTS for voice_id ${voiceId}`);
2023-07-20 19:32:15 +02:00
const response = await doExtrasFetch(
`${this.settings.provider_endpoint}/generate`,
{
method: 'POST',
headers: {
'Content-Type': 'application/json',
2023-12-02 21:06:57 +01:00
'Cache-Control': 'no-cache', // Added this line to disable caching of file so new files are always played - Rolyat 7/7/23
2023-07-20 19:32:15 +02:00
},
body: JSON.stringify({
2023-12-02 19:04:51 +01:00
'text': inputText,
'speaker': voiceId,
2023-12-02 21:06:57 +01:00
'session': 'sillytavern',
}),
},
2023-12-02 20:11:06 +01:00
);
2023-07-20 19:32:15 +02:00
if (!response.ok) {
toastr.error(response.statusText, 'TTS Generation Failed');
throw new Error(`HTTP ${response.status}: ${await response.text()}`);
}
2023-12-02 20:11:06 +01:00
return response;
2023-07-20 19:32:15 +02:00
}
2023-11-28 18:24:26 +01:00
async initSession() {
2023-12-02 19:04:51 +01:00
console.info('Silero TTS: requesting new session');
2023-11-28 18:24:26 +01:00
try {
const response = await doExtrasFetch(
`${this.settings.provider_endpoint}/session`,
{
method: 'POST',
headers: {
'Content-Type': 'application/json',
'Cache-Control': 'no-cache',
},
body: JSON.stringify({
2023-12-02 19:04:51 +01:00
'path': 'sillytavern',
2023-11-28 18:24:26 +01:00
}),
2023-12-02 21:06:57 +01:00
},
2023-12-02 20:11:06 +01:00
);
2023-11-28 18:24:26 +01:00
if (!response.ok && response.status !== 404) {
throw new Error(`HTTP ${response.status}: ${await response.text()}`);
}
2023-11-28 18:24:26 +01:00
} catch (error) {
console.info('Silero TTS: endpoint not available', error);
}
}
2023-07-20 19:32:15 +02:00
// Interface not used by Silero TTS
async fetchTtsFromHistory(history_item_id) {
return Promise.resolve(history_item_id);
}
}