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

209 lines
7.3 KiB
JavaScript
Raw Normal View History

2023-12-02 20:11:06 +01:00
import { getRequestHeaders, callPopup } from '../../../script.js';
2024-01-01 20:31:08 +01:00
import { splitRecursive } from '../../utils.js';
2023-12-02 20:11:06 +01:00
import { getPreviewString, saveTtsProviderSettings } from './index.js';
import { initVoiceMap } from './index.js';
2023-07-20 19:32:15 +02:00
2023-12-02 20:11:06 +01:00
export { NovelTtsProvider };
2023-07-20 19:32:15 +02:00
class NovelTtsProvider {
//########//
// Config //
//########//
2023-12-02 20:11:06 +01:00
settings;
voices = [];
separator = ' . ';
audioElement = document.createElement('audio');
2023-07-20 19:32:15 +02:00
defaultSettings = {
voiceMap: {},
2023-12-02 21:06:57 +01:00
customVoices: [],
2023-12-02 20:11:06 +01:00
};
2023-07-20 19:32:15 +02:00
/**
* Perform any text processing before passing to TTS engine.
* @param {string} text Input text
* @returns {string} Processed text
*/
processText(text) {
// Novel reads tilde as a word. Replace with full stop
text = text.replace(/~/g, '.');
2024-03-17 23:31:28 +01:00
// Novel reads asterisk as a word. Remove it
text = text.replace(/\*/g, '');
return text;
}
2023-07-20 19:32:15 +02:00
get settingsHtml() {
let html = `
2023-08-28 20:46:41 +02:00
<div class="novel_tts_hints">
<div>Use NovelAI's TTS engine.</div>
<div>
The default Voice IDs are only examples. Add custom voices and Novel will create a new random voice for it.
Feel free to try different options!
</div>
<i>Hint: Save an API key in the NovelAI API settings to use it here.</i>
</div>
<label for="tts-novel-custom-voices-add">Custom Voices</label>
2023-08-28 20:46:41 +02:00
<div class="tts_custom_voices">
<select id="tts-novel-custom-voices-select"><select>
2023-08-28 20:46:41 +02:00
<i id="tts-novel-custom-voices-add" class="tts-button fa-solid fa-plus fa-xl success" title="Add"></i>
<i id="tts-novel-custom-voices-delete" class="tts-button fa-solid fa-xmark fa-xl failure" title="Delete"></i>
</div>
`;
2023-07-20 19:32:15 +02:00
return html;
}
// Add a new Novel custom voice to provider
2024-01-01 20:31:08 +01:00
async addCustomVoice() {
2023-12-02 20:11:06 +01:00
const voiceName = await callPopup('<h3>Custom Voice name:</h3>', 'input');
this.settings.customVoices.push(voiceName);
this.populateCustomVoices();
initVoiceMap(); // Update TTS extension voiceMap
saveTtsProviderSettings();
}
// Delete selected custom voice from provider
deleteCustomVoice() {
2023-12-02 19:04:51 +01:00
const selected = $('#tts-novel-custom-voices-select').find(':selected').val();
const voiceIndex = this.settings.customVoices.indexOf(selected);
2023-08-28 20:46:41 +02:00
if (voiceIndex !== -1) {
this.settings.customVoices.splice(voiceIndex, 1);
}
2023-12-02 20:11:06 +01:00
this.populateCustomVoices();
initVoiceMap(); // Update TTS extension voiceMap
saveTtsProviderSettings();
}
// Create the UI dropdown list of voices in provider
2024-01-01 20:31:08 +01:00
populateCustomVoices() {
2023-12-02 20:11:06 +01:00
let voiceSelect = $('#tts-novel-custom-voices-select');
voiceSelect.empty();
this.settings.customVoices.forEach(voice => {
2023-12-02 20:11:06 +01:00
voiceSelect.append(`<option>${voice}</option>`);
});
2023-07-20 19:32:15 +02:00
}
async loadSettings(settings) {
2023-07-20 19:32:15 +02:00
// Populate 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
}
2023-12-02 20:11:06 +01:00
$('#tts-novel-custom-voices-add').on('click', () => (this.addCustomVoice()));
2024-01-01 20:31:08 +01:00
$('#tts-novel-custom-voices-delete').on('click', () => (this.deleteCustomVoice()));
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
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
}
}
2023-12-02 20:11:06 +01:00
this.populateCustomVoices();
await this.checkReady();
console.debug('NovelTTS: Settings loaded');
2023-07-20 19:32:15 +02:00
}
// Perform a simple readiness check by trying to fetch voiceIds
// Doesnt really do much for Novel, not seeing a good way to test this at the moment.
2024-01-01 20:31:08 +01:00
async checkReady() {
2023-12-02 20:11:06 +01:00
await this.fetchTtsVoiceObjects();
}
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
}
//#################//
// TTS Interfaces //
//#################//
async getVoice(voiceName) {
if (!voiceName) {
2023-12-02 20:11:06 +01:00
throw 'TTS Voice name not provided';
2023-07-20 19:32:15 +02:00
}
2023-12-04 18:32:41 +01:00
return { name: voiceName, voice_id: voiceName, lang: 'en-US', preview_url: false };
2023-07-20 19:32:15 +02:00
}
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() {
let voices = [
2023-07-20 19:32:15 +02:00
{ name: 'Ligeia', voice_id: 'Ligeia', lang: 'en-US', preview_url: false },
{ name: 'Aini', voice_id: 'Aini', lang: 'en-US', preview_url: false },
{ name: 'Orea', voice_id: 'Orea', lang: 'en-US', preview_url: false },
{ name: 'Claea', voice_id: 'Claea', lang: 'en-US', preview_url: false },
{ name: 'Lim', voice_id: 'Lim', lang: 'en-US', preview_url: false },
{ name: 'Aurae', voice_id: 'Aurae', lang: 'en-US', preview_url: false },
{ name: 'Naia', voice_id: 'Naia', lang: 'en-US', preview_url: false },
{ name: 'Aulon', voice_id: 'Aulon', lang: 'en-US', preview_url: false },
{ name: 'Elei', voice_id: 'Elei', lang: 'en-US', preview_url: false },
{ name: 'Ogma', voice_id: 'Ogma', lang: 'en-US', preview_url: false },
{ name: 'Raid', voice_id: 'Raid', lang: 'en-US', preview_url: false },
{ name: 'Pega', voice_id: 'Pega', lang: 'en-US', preview_url: false },
{ name: 'Lam', voice_id: 'Lam', lang: 'en-US', preview_url: false },
];
// Add in custom voices to the map
2023-08-28 20:46:41 +02:00
let addVoices = this.settings.customVoices.map(voice =>
2023-12-02 21:06:57 +01:00
({ name: voice, voice_id: voice, lang: 'en-US', preview_url: false }),
2023-12-02 20:11:06 +01:00
);
voices = voices.concat(addVoices);
2023-07-20 19:32:15 +02:00
return voices;
}
async previewTtsVoice(id) {
this.audioElement.pause();
this.audioElement.currentTime = 0;
2023-12-02 20:11:06 +01:00
const text = getPreviewString('en-US');
const response = await this.fetchTtsGeneration(text, id);
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}`);
2023-07-20 19:32:15 +02:00
}
const audio = await response.blob();
const url = URL.createObjectURL(audio);
this.audioElement.src = url;
this.audioElement.play();
2024-05-22 00:36:38 +02:00
URL.revokeObjectURL(url);
2023-07-20 19:32:15 +02:00
}
2024-01-01 20:31:08 +01:00
async* fetchTtsGeneration(inputText, voiceId) {
const MAX_LENGTH = 1000;
2023-12-02 20:11:06 +01:00
console.info(`Generating new TTS for voice_id ${voiceId}`);
2024-01-01 20:31:08 +01:00
const chunks = splitRecursive(inputText, MAX_LENGTH);
for (const chunk of chunks) {
const response = await fetch('/api/novelai/generate-voice',
{
method: 'POST',
headers: getRequestHeaders(),
body: JSON.stringify({
'text': chunk,
'voice': voiceId,
}),
},
);
if (!response.ok) {
toastr.error(response.statusText, 'TTS Generation Failed');
throw new Error(`HTTP ${response.status}: ${await response.text()}`);
}
yield response;
2023-07-20 19:32:15 +02:00
}
}
}