| 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 | export { AzureTtsProvider }; |
| 5 | |
| 6 | class AzureTtsProvider { |
| 7 | //########// |
| 8 | // Config // |
| 9 | //########// |
| 10 | |
| 11 | settings; |
| 12 | voices = []; |
| 13 | separator = ' . '; |
| 14 | audioElement = document.createElement('audio'); |
| 15 | |
| 16 | defaultSettings = { |
| 17 | region: '', |
| 18 | voiceMap: {}, |
| 19 | }; |
| 20 | |
| 21 | get settingsHtml() { |
| 22 | let html = ` |
| 23 | <div class="azure_tts_settings"> |
| 24 | <div class="flex-container alignItemsBaseline"> |
| 25 | <h4 for="azure_tts_key" class="flex1 margin0"> |
| 26 | <a href="https://portal.azure.com/" target="_blank">Azure TTS Key</a> |
| 27 | </h4> |
| 28 | <div id="azure_tts_key" class="menu_button menu_button_icon manage-api-keys" data-key="api_key_azure_tts"> |
| 29 | <i class="fa-solid fa-key"></i> |
| 30 | <span>Click to set</span> |
| 31 | </div> |
| 32 | </div> |
| 33 | <label for="azure_tts_region">Region:</label> |
| 34 | <input id="azure_tts_region" type="text" class="text_pole" placeholder="e.g. westus" /> |
| 35 | <hr> |
| 36 | </div> |
| 37 | `; |
| 38 | return html; |
| 39 | } |
| 40 | |
| 41 | constructor() { |
| 42 | this.handler = async function (/** @type {string} */ key) { |
| 43 | if (key !== SECRET_KEYS.AZURE_TTS) return; |
| 44 | $('#azure_tts_key').toggleClass('success', !!secret_state[SECRET_KEYS.AZURE_TTS]); |
| 45 | await this.onRefreshClick(); |
| 46 | }.bind(this); |
| 47 | } |
| 48 | |
| 49 | dispose() { |
| 50 | [event_types.SECRET_WRITTEN, event_types.SECRET_DELETED, event_types.SECRET_ROTATED].forEach(event => { |
| 51 | eventSource.removeListener(event, this.handler); |
| 52 | }); |
| 53 | } |
| 54 | |
| 55 | onSettingsChange() { |
| 56 | // Update dynamically |
| 57 | this.settings.region = String($('#azure_tts_region').val()); |
| 58 | // Reset voices |
| 59 | this.voices = []; |
| 60 | saveTtsProviderSettings(); |
| 61 | } |
| 62 | |
| 63 | async loadSettings(settings) { |
| 64 | // Populate Provider UI given input settings |
| 65 | if (Object.keys(settings).length == 0) { |
| 66 | console.info('Using default TTS Provider settings'); |
| 67 | } |
| 68 | |
| 69 | // Only accept keys defined in defaultSettings |
| 70 | this.settings = this.defaultSettings; |
| 71 | |
| 72 | for (const key in settings) { |
| 73 | if (key in this.settings) { |
| 74 | this.settings[key] = settings[key]; |
| 75 | } else { |
| 76 | throw `Invalid setting passed to TTS Provider: ${key}`; |
| 77 | } |
| 78 | } |
| 79 | |
| 80 | $('#azure_tts_region').val(this.settings.region).on('input', () => this.onSettingsChange()); |
| 81 | $('#azure_tts_key').toggleClass('success', !!secret_state[SECRET_KEYS.AZURE_TTS]); |
| 82 | [event_types.SECRET_WRITTEN, event_types.SECRET_DELETED, event_types.SECRET_ROTATED].forEach(event => { |
| 83 | eventSource.on(event, this.handler); |
| 84 | }); |
| 85 | |
| 86 | try { |
| 87 | await this.checkReady(); |
| 88 | console.debug('Azure: Settings loaded'); |
| 89 | } catch { |
| 90 | console.debug('Azure: Settings loaded, but not ready'); |
| 91 | } |
| 92 | } |
| 93 | |
| 94 | // Perform a simple readiness check by trying to fetch voiceIds |
| 95 | async checkReady() { |
| 96 | if (secret_state[SECRET_KEYS.AZURE_TTS]) { |
| 97 | await this.fetchTtsVoiceObjects(); |
| 98 | } else { |
| 99 | this.voices = []; |
| 100 | } |
| 101 | } |
| 102 | |
| 103 | async onRefreshClick() { |
| 104 | await this.checkReady(); |
| 105 | } |
| 106 | |
| 107 | //#################// |
| 108 | // TTS Interfaces // |
| 109 | //#################// |
| 110 | |
| 111 | async getVoice(voiceName) { |
| 112 | if (this.voices.length == 0) { |
| 113 | this.voices = await this.fetchTtsVoiceObjects(); |
| 114 | } |
| 115 | const match = this.voices.filter( |
| 116 | voice => voice.name == voiceName, |
| 117 | )[0]; |
| 118 | if (!match) { |
| 119 | throw `TTS Voice name ${voiceName} not found`; |
| 120 | } |
| 121 | return match; |
| 122 | } |
| 123 | |
| 124 | async generateTts(text, voiceId) { |
| 125 | const response = await this.fetchTtsGeneration(text, voiceId); |
| 126 | return response; |
| 127 | } |
| 128 | |
| 129 | //###########// |
| 130 | // API CALLS // |
| 131 | //###########// |
| 132 | async fetchTtsVoiceObjects() { |
| 133 | if (!secret_state[SECRET_KEYS.AZURE_TTS]) { |
| 134 | console.warn('Azure TTS API Key not set'); |
| 135 | return []; |
| 136 | } |
| 137 | |
| 138 | if (!this.settings.region) { |
| 139 | console.warn('Azure TTS region not set'); |
| 140 | return []; |
| 141 | } |
| 142 | |
| 143 | const response = await fetch('/api/azure/list', { |
| 144 | method: 'POST', |
| 145 | headers: getRequestHeaders(), |
| 146 | body: JSON.stringify({ |
| 147 | region: this.settings.region, |
| 148 | }), |
| 149 | }); |
| 150 | |
| 151 | if (!response.ok) { |
| 152 | throw new Error(`HTTP ${response.status}: ${await response.text()}`); |
| 153 | } |
| 154 | let responseJson = await response.json(); |
| 155 | responseJson = responseJson |
| 156 | .sort((a, b) => a.Locale.localeCompare(b.Locale) || a.ShortName.localeCompare(b.ShortName)) |
| 157 | .map(x => ({ name: x.ShortName, voice_id: x.ShortName, preview_url: false, lang: x.Locale })); |
| 158 | return responseJson; |
| 159 | } |
| 160 | |
| 161 | /** |
| 162 | * Preview TTS for a given voice ID. |
| 163 | * @param {string} id Voice ID |
| 164 | */ |
| 165 | async previewTtsVoice(id) { |
| 166 | this.audioElement.pause(); |
| 167 | this.audioElement.currentTime = 0; |
| 168 | const voice = await this.getVoice(id); |
| 169 | const text = getPreviewString(voice.lang); |
| 170 | const response = await this.fetchTtsGeneration(text, id); |
| 171 | if (!response.ok) { |
| 172 | throw new Error(`HTTP ${response.status}: ${await response.text()}`); |
| 173 | } |
| 174 | |
| 175 | const audio = await response.blob(); |
| 176 | const url = URL.createObjectURL(audio); |
| 177 | this.audioElement.src = url; |
| 178 | this.audioElement.play(); |
| 179 | this.audioElement.onended = () => URL.revokeObjectURL(url); |
| 180 | } |
| 181 | |
| 182 | async fetchTtsGeneration(text, voiceId) { |
| 183 | if (!secret_state[SECRET_KEYS.AZURE_TTS]) { |
| 184 | throw new Error('Azure TTS API Key not set'); |
| 185 | } |
| 186 | |
| 187 | if (!this.settings.region) { |
| 188 | throw new Error('Azure TTS region not set'); |
| 189 | } |
| 190 | |
| 191 | const response = await fetch('/api/azure/generate', { |
| 192 | method: 'POST', |
| 193 | headers: getRequestHeaders(), |
| 194 | body: JSON.stringify({ |
| 195 | text: text, |
| 196 | voice: voiceId, |
| 197 | region: this.settings.region, |
| 198 | }), |
| 199 | }); |
| 200 | |
| 201 | if (!response.ok) { |
| 202 | toastr.error(response.statusText, 'TTS Generation Failed'); |
| 203 | throw new Error(`HTTP ${response.status}: ${await response.text()}`); |
| 204 | } |
| 205 | |
| 206 | return response; |
| 207 | } |
| 208 | } |