| 1 | +import { saveTtsProviderSettings } from './index.js'; |
| 2 | +import { getCharacters, getPreviewString } from './index.js'; |
| 3 | + |
| 4 | +export { GptSoVITSAdapterProvider }; |
| 5 | + |
| 6 | +/* |
| 7 | + This file is adapted from gpt-sovits-v2.js. It was created because the original file is no longer maintained. |
| 8 | + Some logic has been optimized and more functionality has been added. |
| 9 | +*/ |
| 10 | + |
| 11 | +class GptSoVITSAdapterProvider { |
| 12 | + settings; |
| 13 | + ready = false; |
| 14 | + voices = []; |
| 15 | + separator = '. '; |
| 16 | + audioElement = document.createElement('audio'); |
| 17 | + /* |
| 18 | + do not modify the text, adapter will handle it |
| 19 | + */ |
| 20 | + processText(text) { |
| 21 | + return text; |
| 22 | + } |
| 23 | + |
| 24 | + audioFormats = ['wav', 'ogg', 'silk', 'mp3', 'flac']; |
| 25 | + |
| 26 | + langKey2LangCode = { |
| 27 | + 'zh': 'zh-CN', |
| 28 | + 'en': 'en-US', |
| 29 | + 'ja': 'ja-JP', |
| 30 | + 'ko': 'ko-KR', |
| 31 | + }; |
| 32 | + |
| 33 | + defaultSettings = { |
| 34 | + provider_endpoint: 'http://localhost:9881', |
| 35 | + format: 'wav', |
| 36 | + lang: 'auto', |
| 37 | + streaming: false, |
| 38 | + text_lang: 'zh', |
| 39 | + media_type: 'auto', |
| 40 | + }; |
| 41 | + |
| 42 | + textLangOptions = [ |
| 43 | + { value: 'zh', label: 'Chinese' }, |
| 44 | + { value: 'en', label: 'English' }, |
| 45 | + { value: 'ja', label: 'Japanese' }, |
| 46 | + { value: 'ko', label: 'Korean' }, |
| 47 | + ]; |
| 48 | + |
| 49 | + mediaTypeOptions = [ |
| 50 | + { value: 'auto', label: 'Auto' }, |
| 51 | + { value: 'wav', label: 'WAV' }, |
| 52 | + { value: 'mp3', label: 'MP3' }, |
| 53 | + { value: 'ogg', label: 'OGG' }, |
| 54 | + { value: 'silk', label: 'SILK' }, |
| 55 | + { value: 'flac', label: 'FLAC' }, |
| 56 | + ]; |
| 57 | + |
| 58 | + _generateOptions(options, currentSetting) { |
| 59 | + return options.map(opt => { |
| 60 | + const isSelected = opt.value === currentSetting ? 'selected' : ''; |
| 61 | + return `<option value="${opt.value}" ${isSelected}>${opt.label}</option>`; |
| 62 | + }).join(''); |
| 63 | + } |
| 64 | + get settingsHtml() { |
| 65 | + const currentSettings = this.settings || this.defaultSettings; |
| 66 | + |
| 67 | + let html = ` |
| 68 | + <label for="gpt_sovits_adapter_tts_endpoint">Provider Endpoint:</label> |
| 69 | + <div class="flex1"> |
| 70 | + <input id="gpt_sovits_adapter_tts_endpoint" type="text" class="text_pole" maxlength="250" height="300" value="${this.defaultSettings.provider_endpoint}"/> |
| 71 | + </div> |
| 72 | + <span>Use <a target="_blank" href="https://github.com/guoql666/GPT-SoVITS_sillytavern_adapter">GPT-SoVITS-adapter</a>.</span><br/> |
| 73 | + <label for="text_lang">Text Lang(Inference text language):</label> |
| 74 | + <select id="text_lang" class="text_pole"> |
| 75 | + ${this._generateOptions(this.textLangOptions, currentSettings.text_lang)} |
| 76 | + </select> |
| 77 | + <label for="media_type">Media Type:</label> |
| 78 | + <select id="media_type" class="text_pole"> |
| 79 | + ${this._generateOptions(this.mediaTypeOptions, currentSettings.media_type)} |
| 80 | + </select> |
| 81 | + <br/> |
| 82 | + `; |
| 83 | + |
| 84 | + return html; |
| 85 | + } |
| 86 | + |
| 87 | + onSettingsChange() { |
| 88 | + // Used when provider settings are updated from UI |
| 89 | + this.settings.provider_endpoint = $('#gpt_sovits_adapter_tts_endpoint').val(); |
| 90 | + this.settings.text_lang = $('#text_lang').val(); |
| 91 | + this.settings.media_type = $('#media_type').val(); |
| 92 | + |
| 93 | + saveTtsProviderSettings(); |
| 94 | + this.changeTTSSettings(); |
| 95 | + } |
| 96 | + |
| 97 | + async loadSettings(settings) { |
| 98 | + // Populate Provider UI given input settings |
| 99 | + if (Object.keys(settings).length == 0) { |
| 100 | + console.info('Using default TTS Provider settings'); |
| 101 | + } |
| 102 | + |
| 103 | + // Only accept keys defined in defaultSettings |
| 104 | + this.settings = this.defaultSettings; |
| 105 | + |
| 106 | + for (const key in settings) { |
| 107 | + if (key in this.settings) { |
| 108 | + this.settings[key] = settings[key]; |
| 109 | + } else { |
| 110 | + console.debug(`Ignoring non-user-configurable setting: ${key}`); |
| 111 | + } |
| 112 | + } |
| 113 | + |
| 114 | + // Set initial values from the settings |
| 115 | + $('#tts_endpoint').val(this.settings.provider_endpoint).on('change', this.onSettingsChange.bind(this)); |
| 116 | + $('#text_lang').val(this.settings.text_lang).on('change', this.onSettingsChange.bind(this)); |
| 117 | + $('#media_type').val(this.settings.media_type).on('change', this.onSettingsChange.bind(this)); |
| 118 | + await this.checkReady(); |
| 119 | + console.info('ITS: Settings loaded'); |
| 120 | + } |
| 121 | + |
| 122 | + // Perform a simple readiness check by trying to fetch voiceIds |
| 123 | + async checkReady() { |
| 124 | + await Promise.allSettled([this.fetchTtsVoiceObjects(), this.changeTTSSettings()]); |
| 125 | + } |
| 126 | + |
| 127 | + async onRefreshClick() { |
| 128 | + return await this.checkReady(); |
| 129 | + } |
| 130 | + |
| 131 | + //#################// |
| 132 | + // TTS Interfaces // |
| 133 | + //#################// |
| 134 | + |
| 135 | + async getVoice(voiceName) { |
| 136 | + if (this.voices.length == 0) { |
| 137 | + this.voices = await this.fetchTtsVoiceObjects(); |
| 138 | + } |
| 139 | + |
| 140 | + const match = this.voices.filter( |
| 141 | + v => v.name == voiceName, |
| 142 | + )[0]; |
| 143 | + if (!match) { |
| 144 | + throw `TTS Voice name ${voiceName} not found`; |
| 145 | + } |
| 146 | + return match; |
| 147 | + } |
| 148 | + |
| 149 | + async generateTts(text, voiceId) { |
| 150 | + const response = await this.fetchTtsGeneration(text, voiceId); |
| 151 | + return response; |
| 152 | + } |
| 153 | + |
| 154 | + //###########// |
| 155 | + // API CALLS // |
| 156 | + //###########// |
| 157 | + async fetchTtsVoiceObjects() { |
| 158 | + const response = await fetch(`${this.settings.provider_endpoint}/speakers`); |
| 159 | + console.info(response); |
| 160 | + |
| 161 | + if (!response.ok) { |
| 162 | + throw new Error(`HTTP ${response.status}: ${await response.json()}`); |
| 163 | + } |
| 164 | + const responseJson = await response.json(); |
| 165 | + this.voices = responseJson; |
| 166 | + return responseJson; |
| 167 | + } |
| 168 | + |
| 169 | + // Each time a parameter is changed, we change the configuration |
| 170 | + async changeTTSSettings() { |
| 171 | + } |
| 172 | + |
| 173 | + /** |
| 174 | + * Preview TTS voice by generating a short sample. |
| 175 | + * @param {string} voiceId Voice ID to preview (model_type&speaker_id)) |
| 176 | + */ |
| 177 | + async previewTtsVoice(voiceId) { |
| 178 | + const langCode = this.langKey2LangCode[this.settings.text_lang] || 'zh-CN'; |
| 179 | + const previewText = getPreviewString(langCode); |
| 180 | + const response = await this.fetchTtsGeneration(previewText, voiceId); |
| 181 | + |
| 182 | + const audio = await response.blob(); |
| 183 | + const url = URL.createObjectURL(audio); |
| 184 | + this.audioElement.src = url; |
| 185 | + this.audioElement.play(); |
| 186 | + this.audioElement.onended = () => URL.revokeObjectURL(url); |
| 187 | + } |
| 188 | + |
| 189 | + /** |
| 190 | + * Fetch TTS generation from the API. |
| 191 | + * @param {string} inputText Text to generate TTS for |
| 192 | + * @param {string} voiceId Voice ID to use (model_type&speaker_id)) |
| 193 | + * @returns {Promise<Response>} Fetch response |
| 194 | + */ |
| 195 | + async fetchTtsGeneration(inputText, voiceId, lang = null, forceNoStreaming = false) { |
| 196 | + console.info(`Generating new TTS for voice_id ${voiceId}`); |
| 197 | + |
| 198 | + const params = { |
| 199 | + text: inputText, |
| 200 | + card_name: getCharacters(false), |
| 201 | + use_st_adapter: true, |
| 202 | + target_voice: voiceId, |
| 203 | + text_lang: this.settings.text_lang, |
| 204 | + text_split_method: 'cut5', |
| 205 | + batch_size: 1, |
| 206 | + media_type: this.settings.media_type, |
| 207 | + streaming_mode: 'true', |
| 208 | + }; |
| 209 | + |
| 210 | + const url = `${this.settings.provider_endpoint}/`; |
| 211 | + |
| 212 | + const response = await fetch( |
| 213 | + url, |
| 214 | + { |
| 215 | + method: 'POST', |
| 216 | + headers: { |
| 217 | + 'Content-Type': 'application/json', |
| 218 | + }, |
| 219 | + body: JSON.stringify(params), |
| 220 | + }, |
| 221 | + ); |
| 222 | + if (!response.ok) { |
| 223 | + toastr.error(response.statusText, 'TTS Generation Failed'); |
| 224 | + throw new Error(`HTTP ${response.status}: ${await response.text()}`); |
| 225 | + } |
| 226 | + return response; |
| 227 | + } |
| 228 | + |
| 229 | + // Interface not used |
| 230 | + async fetchTtsFromHistory(history_item_id) { |
| 231 | + return Promise.resolve(history_item_id); |
| 232 | + } |
| 233 | +} |