| 1 | import { event_types, eventSource, getRequestHeaders } from '../../../script.js'; |
| 2 | import { SECRET_KEYS, secret_state } from '../../secrets.js'; |
| 3 | import { getPreviewString, saveTtsProviderSettings } from './index.js'; |
| 4 | |
| 5 | export { ChutesTtsProvider }; |
| 6 | |
| 7 | class ChutesTtsProvider { |
| 8 | settings; |
| 9 | voices = []; |
| 10 | models = []; |
| 11 | separator = ' . '; |
| 12 | |
| 13 | defaultSettings = { |
| 14 | voiceMap: {}, |
| 15 | model: 'kokoro', |
| 16 | speed: 1, |
| 17 | }; |
| 18 | |
| 19 | get settingsHtml() { |
| 20 | let html = ` |
| 21 | <div class="flex-container alignItemsCenter"> |
| 22 | <div class="flex1">Chutes TTS API</div> |
| 23 | <div id="chutes_tts_key" class="menu_button menu_button_icon manage-api-keys" data-key="api_key_chutes"> |
| 24 | <i class="fa-solid fa-key"></i> |
| 25 | <span>API Key</span> |
| 26 | </div> |
| 27 | </div> |
| 28 | <div class="flex-container flexFlowColumn"> |
| 29 | <div class="flex1"> |
| 30 | <label for="chutes_tts_model">Model</label> |
| 31 | <select id="chutes_tts_model" class="text_pole"></select> |
| 32 | </div> |
| 33 | <div> |
| 34 | <label for="chutes_tts_speed">Speed <span id="chutes_tts_speed_output"></span></label> |
| 35 | <input type="range" id="chutes_tts_speed" value="1" min="0.25" max="3" step="0.05"> |
| 36 | </div> |
| 37 | </div>`; |
| 38 | return html; |
| 39 | } |
| 40 | |
| 41 | constructor() { |
| 42 | this.handler = async function (/** @type {string} */ key) { |
| 43 | if (key !== SECRET_KEYS.CHUTES) return; |
| 44 | $('#chutes_tts_key').toggleClass('success', !!secret_state[SECRET_KEYS.CHUTES]); |
| 45 | await this.onRefreshClick(); |
| 46 | }.bind(this); |
| 47 | } |
| 48 | |
| 49 | dispose() { |
| 50 | [event_types.SECRET_WRITTEN, event_types.SECRET_DELETED, event_types.SECRET_ROTATED].forEach(event => { |
| 51 | eventSource.removeListener(event, this.handler); |
| 52 | }); |
| 53 | } |
| 54 | |
| 55 | onSettingsChange() { |
| 56 | this.settings.model = $('#chutes_tts_model').val(); |
| 57 | this.settings.speed = Number($('#chutes_tts_speed').val()); |
| 58 | saveTtsProviderSettings(); |
| 59 | } |
| 60 | |
| 61 | async loadSettings(settings) { |
| 62 | if (Object.keys(settings).length === 0) { |
| 63 | Object.assign(settings, this.defaultSettings); |
| 64 | } |
| 65 | |
| 66 | this.settings = settings; |
| 67 | |
| 68 | if (!this.settings.voiceMap) { |
| 69 | this.settings.voiceMap = {}; |
| 70 | } |
| 71 | |
| 72 | // Update UI |
| 73 | $('#chutes_tts_model').val(this.settings.model); |
| 74 | $('#chutes_tts_speed').val(this.settings.speed); |
| 75 | $('#chutes_tts_speed_output').text(this.settings.speed); |
| 76 | |
| 77 | $('#chutes_tts_key').toggleClass('success', !!secret_state[SECRET_KEYS.CHUTES]); |
| 78 | [event_types.SECRET_WRITTEN, event_types.SECRET_DELETED, event_types.SECRET_ROTATED].forEach(event => { |
| 79 | eventSource.on(event, this.handler); |
| 80 | }); |
| 81 | |
| 82 | await this.checkReady(); |
| 83 | |
| 84 | $('#chutes_tts_model').on('change', () => this.onSettingsChange()); |
| 85 | $('#chutes_tts_speed').on('input', () => { |
| 86 | const value = $('#chutes_tts_speed').val(); |
| 87 | $('#chutes_tts_speed_output').text(String(value)); |
| 88 | this.onSettingsChange(); |
| 89 | }); |
| 90 | } |
| 91 | |
| 92 | async checkReady() { |
| 93 | await this.updateModels(); |
| 94 | if (this.models.length === 0) { |
| 95 | // No models available |
| 96 | } |
| 97 | await this.updateVoices(); |
| 98 | } |
| 99 | |
| 100 | async onRefreshClick() { |
| 101 | return await this.checkReady(); |
| 102 | } |
| 103 | |
| 104 | async updateModels() { |
| 105 | // For Chutes TTS, we always use the Kokoro model currently. |
| 106 | this.models = ['kokoro']; |
| 107 | |
| 108 | $('#chutes_tts_model').empty(); |
| 109 | $('#chutes_tts_model').append($('<option>').val('kokoro').text('Kokoro')); |
| 110 | $('#chutes_tts_model').val('kokoro'); |
| 111 | |
| 112 | this.settings.model = 'kokoro'; |
| 113 | } |
| 114 | |
| 115 | async updateVoices() { |
| 116 | // Kokoro voices list |
| 117 | const kokoroVoices = [ |
| 118 | { id: 'af_alloy', name: 'Alloy (Female)', lang: 'en-US' }, |
| 119 | { id: 'af_aoede', name: 'Aoede (Female)', lang: 'en-US' }, |
| 120 | { id: 'af_bella', name: 'Bella (Female)', lang: 'en-US' }, |
| 121 | { id: 'af_heart', name: 'Heart (Female) - Default', lang: 'en-US' }, |
| 122 | { id: 'af_jessica', name: 'Jessica (Female)', lang: 'en-US' }, |
| 123 | { id: 'af_kore', name: 'Kore (Female)', lang: 'en-US' }, |
| 124 | { id: 'af_nicole', name: 'Nicole (Female)', lang: 'en-US' }, |
| 125 | { id: 'af_nova', name: 'Nova (Female)', lang: 'en-US' }, |
| 126 | { id: 'af_river', name: 'River (Female)', lang: 'en-US' }, |
| 127 | { id: 'af_sarah', name: 'Sarah (Female)', lang: 'en-US' }, |
| 128 | { id: 'af_sky', name: 'Sky (Female)', lang: 'en-US' }, |
| 129 | { id: 'am_adam', name: 'Adam (Male)', lang: 'en-US' }, |
| 130 | { id: 'am_echo', name: 'Echo (Male)', lang: 'en-US' }, |
| 131 | { id: 'am_eric', name: 'Eric (Male)', lang: 'en-US' }, |
| 132 | { id: 'am_fenrir', name: 'Fenrir (Male)', lang: 'en-US' }, |
| 133 | { id: 'am_liam', name: 'Liam (Male)', lang: 'en-US' }, |
| 134 | { id: 'am_michael', name: 'Michael (Male)', lang: 'en-US' }, |
| 135 | { id: 'am_onyx', name: 'Onyx (Male)', lang: 'en-US' }, |
| 136 | { id: 'am_puck', name: 'Puck (Male)', lang: 'en-US' }, |
| 137 | { id: 'am_santa', name: 'Santa (Male)', lang: 'en-US' }, |
| 138 | { id: 'bf_alice', name: 'Alice (British Female)', lang: 'en-GB' }, |
| 139 | { id: 'bf_emma', name: 'Emma (British Female)', lang: 'en-GB' }, |
| 140 | { id: 'bf_isabella', name: 'Isabella (British Female)', lang: 'en-GB' }, |
| 141 | { id: 'bf_lily', name: 'Lily (British Female)', lang: 'en-GB' }, |
| 142 | { id: 'bm_daniel', name: 'Daniel (British Male)', lang: 'en-GB' }, |
| 143 | { id: 'bm_fable', name: 'Fable (British Male)', lang: 'en-GB' }, |
| 144 | { id: 'bm_george', name: 'George (British Male)', lang: 'en-GB' }, |
| 145 | { id: 'bm_lewis', name: 'Lewis (British Male)', lang: 'en-GB' }, |
| 146 | { id: 'ef_dora', name: 'Dora (European Female)', lang: 'es-ES' }, |
| 147 | { id: 'em_alex', name: 'Alex (European Male)', lang: 'es-ES' }, |
| 148 | { id: 'em_santa', name: 'Santa (European Male)', lang: 'es-ES' }, |
| 149 | { id: 'ff_siwis', name: 'Siwis (French Female)', lang: 'fr-FR' }, |
| 150 | { id: 'hf_alpha', name: 'Alpha (Hindi Female)', lang: 'hi-IN' }, |
| 151 | { id: 'hf_beta', name: 'Beta (Hindi Female)', lang: 'hi-IN' }, |
| 152 | { id: 'hm_omega', name: 'Omega (Hindi Male)', lang: 'hi-IN' }, |
| 153 | { id: 'hm_psi', name: 'Psi (Hindi Male)', lang: 'hi-IN' }, |
| 154 | { id: 'if_sara', name: 'Sara (Italian Female)', lang: 'it-IT' }, |
| 155 | { id: 'im_nicola', name: 'Nicola (Italian Male)', lang: 'it-IT' }, |
| 156 | { id: 'jf_alpha', name: 'Alpha (Japanese Female)', lang: 'ja-JP' }, |
| 157 | { id: 'jf_gongitsune', name: 'Gongitsune (Japanese Female)', lang: 'ja-JP' }, |
| 158 | { id: 'jf_nezumi', name: 'Nezumi (Japanese Female)', lang: 'ja-JP' }, |
| 159 | { id: 'jf_tebukuro', name: 'Tebukuro (Japanese Female)', lang: 'ja-JP' }, |
| 160 | { id: 'jm_kumo', name: 'Kumo (Japanese Male)', lang: 'ja-JP' }, |
| 161 | { id: 'pf_dora', name: 'Dora (Portuguese Female)', lang: 'pt-PT' }, |
| 162 | { id: 'pm_alex', name: 'Alex (Portuguese Male)', lang: 'pt-PT' }, |
| 163 | { id: 'pm_santa', name: 'Santa (Portuguese Male)', lang: 'pt-PT' }, |
| 164 | { id: 'zf_xiaobei', name: 'Xiaobei (Chinese Female)', lang: 'zh-CN' }, |
| 165 | { id: 'zf_xiaoni', name: 'Xiaoni (Chinese Female)', lang: 'zh-CN' }, |
| 166 | { id: 'zf_xiaoxiao', name: 'Xiaoxiao (Chinese Female)', lang: 'zh-CN' }, |
| 167 | { id: 'zf_xiaoyi', name: 'Xiaoyi (Chinese Female)', lang: 'zh-CN' }, |
| 168 | { id: 'zm_yunjian', name: 'Yunjian (Chinese Male)', lang: 'zh-CN' }, |
| 169 | { id: 'zm_yunxi', name: 'Yunxi (Chinese Male)', lang: 'zh-CN' }, |
| 170 | { id: 'zm_yunxia', name: 'Yunxia (Chinese Male)', lang: 'zh-CN' }, |
| 171 | { id: 'zm_yunyang', name: 'Yunyang (Chinese Male)', lang: 'zh-CN' }, |
| 172 | ]; |
| 173 | |
| 174 | this.voices = kokoroVoices.map(v => ({ |
| 175 | name: v.name, |
| 176 | voice_id: v.id, |
| 177 | lang: v.lang, |
| 178 | })); |
| 179 | } |
| 180 | |
| 181 | async getVoice(voiceName) { |
| 182 | if (this.voices.length === 0) { |
| 183 | await this.updateVoices(); |
| 184 | } |
| 185 | const voice = this.voices.find(v => v.name === voiceName || v.voice_id === voiceName); |
| 186 | return voice || this.voices.find(v => v.voice_id === 'af_heart'); |
| 187 | } |
| 188 | |
| 189 | async generateTts(text, voiceId) { |
| 190 | const response = await this.fetchTtsGeneration(text, voiceId); |
| 191 | return response; |
| 192 | } |
| 193 | |
| 194 | async fetchTtsGeneration(text, voiceId) { |
| 195 | const apiKey = secret_state[SECRET_KEYS.CHUTES]; |
| 196 | |
| 197 | if (!apiKey) { |
| 198 | throw new Error('No Chutes API key found'); |
| 199 | } |
| 200 | |
| 201 | const response = await fetch('/api/openai/chutes/generate-voice', { |
| 202 | method: 'POST', |
| 203 | headers: getRequestHeaders(), |
| 204 | body: JSON.stringify({ |
| 205 | input: text, |
| 206 | voice: voiceId || 'af_heart', |
| 207 | speed: this.settings.speed || 1, |
| 208 | }), |
| 209 | }); |
| 210 | |
| 211 | if (!response.ok) { |
| 212 | const error = await response.text(); |
| 213 | throw new Error(`Chutes TTS failed: ${error}`); |
| 214 | } |
| 215 | |
| 216 | return response; |
| 217 | } |
| 218 | |
| 219 | async fetchTtsVoiceObjects() { |
| 220 | if (this.voices.length === 0) { |
| 221 | await this.updateVoices(); |
| 222 | } |
| 223 | |
| 224 | const voiceIds = this.voices |
| 225 | .map(voice => ({ name: voice.name, voice_id: voice.voice_id, preview_url: false })); |
| 226 | return voiceIds; |
| 227 | } |
| 228 | |
| 229 | async previewTtsVoice(voiceId) { |
| 230 | const text = getPreviewString(voiceId); |
| 231 | await this.generateTts(text, voiceId); |
| 232 | } |
| 233 | } |