Switch branch
feat/multi-feature-pack maindefault
View all branches
Blame Raw
Cohee · 51ad27fb · · 1622 lines (58.0 KB)
5 contributors 
1import { cancelTtsPlay, eventSource, event_types, getCurrentChatId, isStreamingEnabled, name2, saveSettingsDebounced, substituteParams } from '../../../script.js';


2import { ModuleWorkerWrapper, extension_settings, getContext, renderExtensionTemplateAsync } from '../../extensions.js';


3import { delay, escapeRegex, getBase64Async, getStringHash, onlyUnique, regexFromString } from '../../utils.js';


4import { accountStorage } from '../../util/AccountStorage.js';


5import { EdgeTtsProvider } from './edge.js';


6import { ElevenLabsTtsProvider } from './elevenlabs.js';


7import { SileroTtsProvider } from './silerotts.js';


8import { GptSovitsV2Provider } from './gpt-sovits-v2.js';


9import { GptSoVITSAdapterProvider } from './gpt-sovits-adapter.js';


10import { CoquiTtsProvider } from './coqui.js';


11import { SystemTtsProvider } from './system.js';


12import { NovelTtsProvider } from './novel.js';


13import { power_user } from '../../power-user.js';


14import { OpenAITtsProvider } from './openai.js';


15import { OpenAICompatibleTtsProvider } from './openai-compatible.js';


16import { XTTSTtsProvider } from './xtts.js';


17import { VITSTtsProvider } from './vits.js';


18import { GSVITtsProvider } from './gsvi.js';


19import { SBVits2TtsProvider } from './sbvits2.js';


20import { AllTalkTtsProvider } from './alltalk.js';


21import { CosyVoiceProvider } from './cosyvoice.js';


22import { SpeechT5TtsProvider } from './speecht5.js';


23import { AzureTtsProvider } from './azure.js';


24import { SlashCommandParser } from '../../slash-commands/SlashCommandParser.js';


25import { SlashCommand } from '../../slash-commands/SlashCommand.js';


26import { ARGUMENT_TYPE, SlashCommandArgument, SlashCommandNamedArgument } from '../../slash-commands/SlashCommandArgument.js';


27import { debounce_timeout } from '../../constants.js';


28import { SlashCommandEnumValue, enumTypes } from '../../slash-commands/SlashCommandEnumValue.js';


29import { enumIcons } from '../../slash-commands/SlashCommandCommonEnumsProvider.js';


30import { POPUP_TYPE, callGenericPopup } from '../../popup.js';


31import { GoogleTranslateTtsProvider } from './google-translate.js';


32import { GoogleNativeTtsProvider } from './google-native.js';


33import { ChatterboxTtsProvider } from './chatterbox.js';


34import { KokoroTtsProvider } from './kokoro.js';


35import { TtsWebuiProvider } from './tts-webui.js';


36import { PollinationsTtsProvider } from './pollinations.js';


37import { MiniMaxTtsProvider } from './minimax.js';


38import { ElectronHubTtsProvider } from './electronhub.js';


39import { ChutesTtsProvider } from './chutes.js';


40import { VolcengineTtsProvider } from './volcengine.js';


41import { applyLocale, t } from '/scripts/i18n.js';


42


43const UPDATE_INTERVAL = 1000;


44const wrapper = new ModuleWorkerWrapper(moduleWorker);


45


46let voiceMapEntries = [];


47let voiceMap = {}; // {charName:voiceid, charName2:voiceid2}


48let lastChatId = null;


49let lastMessage = null;


50let lastMessageHash = null;


51let periodicMessageGenerationTimer = null;


52let lastPositionOfParagraphEnd = -1;


53let currentInitVoiceMapPromise = null;


54


55const DEFAULT_VOICE_MARKER = '[Default Voice]';


56const DISABLED_VOICE_MARKER = 'disabled';


57


58export function getPreviewString(lang) {


59    const previewStrings = {


60        'en-US': 'The quick brown fox jumps over the lazy dog',


61        'en-GB': 'Sphinx of black quartz, judge my vow',


62        'fr-FR': 'Portez ce vieux whisky au juge blond qui fume',


63        'de-DE': 'Victor jagt zwölf Boxkämpfer quer über den großen Sylter Deich',


64        'it-IT': 'Pranzo d\'acqua fa volti sghembi',


65        'es-ES': 'Quiere la boca exhausta vid, kiwi, piña y fugaz jamón',


66        'es-MX': 'Fabio me exige, sin tapujos, que añada cerveza al whisky',


67        'ru-RU': 'В чащах юга жил бы цитрус? Да, но фальшивый экземпляр!',


68        'pt-BR': 'Vejo xá gritando que fez show sem playback.',


69        'pt-PR': 'Todo pajé vulgar faz boquinha sexy com kiwi.',


70        'uk-UA': 'Фабрикуймо гідність, лящім їжею, ґав хапаймо, з\'єднавці чаш!',


71        'pl-PL': 'Pchnąć w tę łódź jeża lub ośm skrzyń fig',


72        'cs-CZ': 'Příliš žluťoučký kůň úpěl ďábelské ódy',


73        'sk-SK': 'Vyhŕňme si rukávy a vyprážajme čínske ryžové cestoviny',


74        'hu-HU': 'Árvíztűrő tükörfúrógép',


75        'tr-TR': 'Pijamalı hasta yağız şoföre çabucak güvendi',


76        'nl-NL': 'De waard heeft een kalfje en een pinkje opgegeten',


77        'sv-SE': 'Yxskaftbud, ge vårbygd, zinkqvarn',


78        'da-DK': 'Quizdeltagerne spiste jordbær med fløde, mens cirkusklovnen Walther spillede på xylofon',


79        'ja-JP': 'いろはにほへと　ちりぬるを　わかよたれそ　つねならむ　うゐのおくやま　けふこえて　あさきゆめみし　ゑひもせす',


80        'ko-KR': '가나다라마바사아자차카타파하',


81        'zh-CN': '我能吞下玻璃而不伤身体',


82        'ro-RO': 'Muzicologă în bej vând whisky și tequila, preț fix',


83        'bg-BG': 'Щъркелите се разпръснаха по цялото небе',


84        'el-GR': 'Ταχίστη αλώπηξ βαφής ψημένη γη, δρασκελίζει υπέρ νωθρού κυνός',


85        'fi-FI': 'Voi veljet, miksi juuri teille myin nämä vehkeet?',


86        'he-IL': 'הקצינים צעקו: "כל הכבוד לצבא הצבאות!"',


87        'id-ID': 'Jangkrik itu memang enak, apalagi kalau digoreng',


88        'ms-MY': 'Muzik penyanyi wanita itu menggambarkan kehidupan yang penuh dengan duka nestapa',


89        'th-TH': 'เป็นไงบ้างครับ ผมชอบกินข้าวผัดกระเพราหมูกรอบ',


90        'vi-VN': 'Cô bé quàng khăn đỏ đang ngồi trên bãi cỏ xanh',


91        'ar-SA': 'أَبْجَدِيَّة عَرَبِيَّة',


92        'hi-IN': 'श्वेता ने श्वेता के श्वेते हाथों में श्वेता का श्वेता चावल पकड़ा',


93    };


94    const fallbackPreview = 'Neque porro quisquam est qui dolorem ipsum quia dolor sit amet';


95


96    return previewStrings[lang] ?? fallbackPreview;


97}


98


99/**


100 * Registers a TTS provider.


101 * @param {string} name Name of the TTS provider to register.


102 * @param {function} provider Provider class.


103 */


104export function registerTtsProvider(name, provider) {


105    if (!name || typeof name !== 'string') {


106        throw new Error(`TTS provider name ${name} is not a valid string.`);


107    }


108    if (!provider || typeof provider !== 'function') {


109        throw new Error(`TTS provider ${name} is not a valid provider class.`);


110    }


111    if (ttsProviders[name]) {


112        throw new Error(`TTS provider ${name} is already registered.`);


113    }


114    ttsProviders[name] = provider;


115    console.info(`Registered TTS provider: ${name}`);


116    $('#tts_provider').append($('<option />').val(name).text(name));


117


118    // Load if it was previously selected


119    if (extension_settings.tts.currentProvider === name) {


120        loadTtsProvider(name);


121    }


122}


123


124const ttsProviders = {


125    AllTalk: AllTalkTtsProvider,


126    Azure: AzureTtsProvider,


127    Chatterbox: ChatterboxTtsProvider,


128    Chutes: ChutesTtsProvider,


129    Coqui: CoquiTtsProvider,


130    'CosyVoice (Unofficial)': CosyVoiceProvider,


131    Edge: EdgeTtsProvider,


132    ElevenLabs: ElevenLabsTtsProvider,


133    'Electron Hub': ElectronHubTtsProvider,


134    'Google Translate': GoogleTranslateTtsProvider,


135    'Google Gemini TTS': GoogleNativeTtsProvider,


136    GSVI: GSVITtsProvider,


137    'GPT-SoVITS-Adapter': GptSoVITSAdapterProvider,


138    'GPT-SoVITS-V2 (Unofficial)': GptSovitsV2Provider,


139    Kokoro: KokoroTtsProvider,


140    MiniMax: MiniMaxTtsProvider,


141    Novel: NovelTtsProvider,


142    OpenAI: OpenAITtsProvider,


143    'OpenAI Compatible': OpenAICompatibleTtsProvider,


144    Pollinations: PollinationsTtsProvider,


145    SBVits2: SBVits2TtsProvider,


146    Silero: SileroTtsProvider,


147    SpeechT5: SpeechT5TtsProvider,


148    System: SystemTtsProvider,


149    'TTS WebUI': TtsWebuiProvider,


150    VITS: VITSTtsProvider,


151    XTTSv2: XTTSTtsProvider,


152    Volcengine: VolcengineTtsProvider,


153};


154let ttsProvider;


155let ttsProviderName;


156


157


158async function onNarrateOneMessage() {


159    audioElement.src = '/sounds/silence.mp3';


160    const context = getContext();


161    const id = $(this).closest('.mes').attr('mesid');


162    const message = context.chat[id];


163


164    if (!message) {


165        return;


166    }


167


168    resetTtsPlayback();


169    processAndQueueTtsMessage(message, Number(id), { manual: true });


170    moduleWorker();


171}


172


173async function onNarrateText(args, text) {


174    if (!text) {


175        return '';


176    }


177


178    audioElement.src = '/sounds/silence.mp3';


179


180    // To load all characters in the voice map, set unrestricted to true


181    await initVoiceMap(true);


182


183    const baseName = args?.voice || name2;


184    const name = (baseName === 'SillyTavern System' ? DEFAULT_VOICE_MARKER : baseName) || DEFAULT_VOICE_MARKER;


185


186    const voiceMapEntry = voiceMap[name] === DEFAULT_VOICE_MARKER


187        ? voiceMap[DEFAULT_VOICE_MARKER]


188        : voiceMap[name];


189


190    if (voiceMapEntry === DISABLED_VOICE_MARKER) {


191        toastr.info(`TTS voice for ${name} is disabled.`);


192        await initVoiceMap(false);


193        return;


194    }


195


196    if (!voiceMapEntry) {


197        toastr.info(`Specified voice for ${name} was not found. Check the TTS extension settings.`);


198        await initVoiceMap(false);


199        return;


200    }


201


202    resetTtsPlayback();


203    processAndQueueTtsMessage({ mes: text, name: name }, null, { manual: true });


204    await moduleWorker();


205


206    // Return back to the chat voices


207    await initVoiceMap(false);


208    return '';


209}


210


211async function moduleWorker() {


212    if (!extension_settings.tts.enabled) {


213        return;


214    }


215


216    processTtsQueue();


217    processAudioJobQueue();


218    updateUiAudioPlayState();


219}


220


221function resetTtsPlayback() {


222    // Stop system TTS utterance


223    cancelTtsPlay();


224


225    // Clear currently processing jobs


226    currentTtsJob = null;


227    currentAudioJob = null;


228


229    // Reset audio element


230    audioElement.currentTime = 0;


231    audioElement.src = '';


232


233    // Clear any queue items


234    ttsJobQueue.splice(0, ttsJobQueue.length);


235    audioJobQueue.splice(0, audioJobQueue.length);


236


237    // Set audio ready to process again


238    audioQueueProcessorReady = true;


239}


240


241function isTtsProcessing() {


242    let processing = false;


243


244    // Check job queues


245    if (ttsJobQueue.length > 0 || audioJobQueue.length > 0) {


246        processing = true;


247    }


248    // Check current jobs


249    if (currentTtsJob != null || currentAudioJob != null) {


250        processing = true;


251    }


252    return processing;


253}


254


255/**


256 * @typedef {ChatMessage & { id?: number, manual?: boolean, segmentText?: string, segmentType?: string }} TtsMessage


257 */


258


259/**


260 * Clones a message, attaches the given message ID, then splits by paragraphs


261 * (if enabled) and adds each part to the TTS job queue.


262 * @param {ChatMessage} message - The message object to be processed.


263 * @param {number|null} [messageId=null] - The chat message index to associate with TTS events.


264 * @param {object} [options={}] - Additional options for processing.


265 * @param {boolean} [options.manual=false] - Whether this TTS job was manually triggered (e.g., from the UI) rather than automatically from a new chat message.


266 * @returns {void}


267 */


268function processAndQueueTtsMessage(message, messageId = null, { manual = false } = {}) {


269    /** @type {TtsMessage} */


270    const clone = structuredClone(message);


271    clone.id = messageId ?? null;


272    clone.manual = manual ?? false;


273


274    if (!extension_settings.tts.narrate_by_paragraphs) {


275        ttsJobQueue.push(clone);


276        return;


277    }


278


279    const lines = clone.mes.split('\n');


280


281    for (let i = 0; i < lines.length; i++) {


282        const line = lines[i];


283


284        if (line.length === 0) {


285            continue;


286        }


287


288        ttsJobQueue.push(


289            Object.assign({}, clone, {


290                mes: line,


291            }),


292        );


293    }


294}


295


296function debugTtsPlayback() {


297    console.log(JSON.stringify(


298        {


299            'ttsProviderName': ttsProviderName,


300            'voiceMap': voiceMap,


301            'audioPaused': audioPaused,


302            'audioJobQueue': audioJobQueue,


303            'currentAudioJob': currentAudioJob,


304            'audioQueueProcessorReady': audioQueueProcessorReady,


305            'ttsJobQueue': ttsJobQueue,


306            'currentTtsJob': currentTtsJob,


307            'ttsConfig': extension_settings.tts,


308        },


309    ));


310}


311globalThis.debugTtsPlayback = debugTtsPlayback;


312


313//##################//


314//   Audio Control  //


315//##################//


316


317let audioElement = new Audio();


318audioElement.id = 'tts_audio';


319audioElement.autoplay = true;


320


321/**


322 * @type AudioJob[] Audio job queue


323 * @typedef {{audioBlob: Blob | string, char: string}} AudioJob Audio job object


324 */


325const audioJobQueue = [];


326/**


327 * @type AudioJob Current audio job


328 */


329let currentAudioJob;


330let audioPaused = false;


331let audioQueueProcessorReady = true;


332


333/**


334 * Play audio data from audio job object.


335 * @param {AudioJob} audioJob Audio job object


336 * @returns {Promise<void>} Promise that resolves when audio playback is started


337 */


338async function playAudioData(audioJob) {


339    const { audioBlob, char } = audioJob;


340    // Since current audio job can be cancelled, don't playback if it is null


341    if (currentAudioJob == null) {


342        console.log('Cancelled TTS playback because currentAudioJob was null');


343    }


344    if (audioBlob instanceof Blob) {


345        const srcUrl = await getBase64Async(audioBlob);


346


347        // VRM lip sync


348        if (extension_settings.vrm?.enabled && typeof globalThis.vrmLipSync === 'function') {


349            await globalThis.vrmLipSync(audioBlob, char);


350        }


351


352        audioElement.src = srcUrl;


353    } else if (typeof audioBlob === 'string') {


354        audioElement.src = audioBlob;


355    } else {


356        throw `TTS received invalid audio data type ${typeof audioBlob}`;


357    }


358    audioElement.addEventListener('ended', completeCurrentAudioJob);


359    audioElement.addEventListener('canplay', () => {


360        console.debug('Starting TTS playback');


361        audioElement.playbackRate = extension_settings.tts.playback_rate;


362        audioElement.play();


363    });


364}


365


366globalThis.tts_preview = function (id) {


367    const audio = document.getElementById(id);


368


369    if (audio instanceof HTMLAudioElement && !$(audio).data('disabled')) {


370        audio.play();


371    } else {


372        ttsProvider.previewTtsVoice(id);


373    }


374};


375


376async function onTtsVoicesClick() {


377    let popupText = '';


378


379    try {


380        const voiceIds = await ttsProvider.fetchTtsVoiceObjects();


381


382        for (const voice of voiceIds) {


383            popupText += `


384            <div class="voice_preview">


385                <span class="voice_lang">${voice.lang || ''}</span>


386                <b class="voice_name">${voice.name}</b>


387                <i onclick="tts_preview('${voice.voice_id}')" class="fa-solid fa-play"></i>


388            </div>`;


389            if (voice.preview_url) {


390                popupText += `<audio id="${voice.voice_id}" src="${voice.preview_url}" data-disabled="${voice.preview_url == false}"></audio>`;


391            }


392        }


393    } catch {


394        popupText = 'Could not load voices list. Check your API key.';


395    }


396


397    callGenericPopup(popupText, POPUP_TYPE.TEXT, '', { allowVerticalScrolling: true });


398}


399


400function updateUiAudioPlayState() {


401    if (extension_settings.tts.enabled == true) {


402        $('#ttsExtensionMenuItem').show();


403        let img;


404        // Give user feedback that TTS is active by setting the stop icon if processing or playing


405        if (!audioElement.paused || isTtsProcessing()) {


406            img = 'fa-solid fa-stop-circle extensionsMenuExtensionButton';


407        } else {


408            img = 'fa-solid fa-circle-play extensionsMenuExtensionButton';


409        }


410        $('#tts_media_control').attr('class', img);


411    } else {


412        $('#ttsExtensionMenuItem').hide();


413    }


414}


415


416function onAudioControlClicked() {


417    audioElement.src = '/sounds/silence.mp3';


418    let context = getContext();


419    // Not pausing, doing a full stop to anything TTS is doing. Better UX as pause is not as useful


420    if (!audioElement.paused || isTtsProcessing()) {


421        resetTtsPlayback();


422    } else if (context?.chat?.length > 0) {


423        // Default play behavior if not processing or playing is to play the last message.


424        const id = context.chat.length - 1;


425        processAndQueueTtsMessage(context.chat[id], id, { manual: true });


426    }


427    updateUiAudioPlayState();


428}


429


430function addAudioControl() {


431    $('#tts_wand_container').append(applyLocale(`


432        <div id="ttsExtensionMenuItem" class="list-group-item flex-container flexGap5">


433            <div id="tts_media_control" class="extensionsMenuExtensionButton "/></div>


434            <span data-i18n="TTS Playback">TTS Playback</span>


435        </div>`));


436    $('#tts_wand_container').append(applyLocale(`


437        <div id="ttsExtensionNarrateAll" class="list-group-item flex-container flexGap5">


438            <div class="extensionsMenuExtensionButton fa-solid fa-radio"></div>


439            <span data-i18n="Narrate All Chat">Narrate All Chat</span>


440        </div>`));


441    $('#ttsExtensionMenuItem').attr('title', t`TTS play/pause`).attr('data-i18n', '[title]TTS play/pause').on('click', onAudioControlClicked);


442    $('#ttsExtensionNarrateAll').attr('title', t`Narrate all messages in the current chat. Includes user messages, excludes hidden comments.`).attr('data-i18n', '[title]Narrate all messages in the current chat. Includes user messages, excludes hidden comments.').on('click', playFullConversation);


443    updateUiAudioPlayState();


444}


445


446function completeCurrentAudioJob() {


447    audioQueueProcessorReady = true;


448    currentAudioJob = null;


449    // updateUiPlayState();


450    wrapper.update();


451}


452


453/**


454 * Accepts an HTTP response containing audio/mpeg data, and puts the data as a Blob() on the queue for playback


455 * @param {Response} response


456 * @param {string} char


457 * @returns {Promise<{audioBlob: Blob|string, mimeType: string}>}


458 */


459async function addAudioJob(response, char) {


460    let audioBlob, mimeType;


461    if (typeof response === 'string') {


462        audioBlob = response;


463        mimeType = '';


464    } else {


465        audioBlob = await response.blob();


466        if (!audioBlob.type.startsWith('audio/')) {


467            throw `TTS received HTTP response with invalid data format. Expecting audio/*, got ${audioBlob.type}`;


468        }


469        mimeType = audioBlob.type;


470    }


471    audioJobQueue.push({ audioBlob, char });


472    console.debug('Pushed audio job to queue.');


473    return { audioBlob, mimeType };


474}


475


476async function processAudioJobQueue() {


477    // Nothing to do, audio not completed, or audio paused - stop processing.


478    if (audioJobQueue.length == 0 || !audioQueueProcessorReady || audioPaused) {


479        return;


480    }


481    try {


482        audioQueueProcessorReady = false;


483        currentAudioJob = audioJobQueue.shift();


484        playAudioData(currentAudioJob);


485    } catch (error) {


486        toastr.error(error.toString());


487        console.error(error);


488        audioQueueProcessorReady = true;


489    }


490}


491


492//################//


493//  TTS Control   //


494//################//


495


496/** @type {TtsMessage[]} */


497const ttsJobQueue = [];


498/** @type {TtsMessage|null} */


499let currentTtsJob = null; // Null if nothing is currently being processed


500


501function completeTtsJob() {


502    console.info(`Current TTS job for ${currentTtsJob?.name} completed.`);


503    currentTtsJob = null;


504}


505


506async function tts(text, voiceId, char, voiceMapKey = null) {


507    const messageId = currentTtsJob?.id ?? null;


508


509    await eventSource.emit(event_types.TTS_JOB_STARTED, { messageId, characterName: char, text, voiceId });


510


511    async function processResponse(response) {


512        // RVC injection


513        if (typeof globalThis.rvcVoiceConversion === 'function' && extension_settings.rvc.enabled)


514            response = await globalThis.rvcVoiceConversion(response, char, text);


515


516        const audioResult = await addAudioJob(response, char);


517        const eventData = { messageId, characterName: char, text, audio: audioResult.audioBlob, mimeType: audioResult.mimeType };


518        await eventSource.emit(event_types.TTS_AUDIO_READY, eventData);


519    }


520


521    // voiceMapKey can also include segment qualifiers, e.g. '{char} ("Quotes")'


522    let response = await ttsProvider.generateTts(text, voiceId, voiceMapKey);


523


524    // If async generator, process every chunk as it comes in


525    if (typeof response[Symbol.asyncIterator] === 'function') {


526        for await (const chunk of response) {


527            await processResponse(chunk);


528        }


529    } else {


530        await processResponse(response);


531    }


532


533    await eventSource.emit(event_types.TTS_JOB_COMPLETE, { messageId, characterName: char });


534    completeTtsJob();


535}


536


537function parseMessageSegments(text) {


538    if (!extension_settings.tts.multi_voice_enabled) {


539        return [{ type: 'other', text: text }];


540    }


541


542    const segments = [];


543    const segmentRegex = /(\*[^*]*?\*)|(".*?")|(\u201C.*?\u201D)|(\u00AB.*?\u00BB)|(\u300C.*?\u300D)|(\u300E.*?\u300F)|(\uFF02.*?\uFF02)/gim;


544    let lastIndex = 0;


545    let match;


546


547    segmentRegex.lastIndex = 0;


548


549    while ((match = segmentRegex.exec(text)) !== null) {


550        // Add other text before this match


551        if (match.index > lastIndex) {


552            const otherText = text.substring(lastIndex, match.index).trim();


553            if (otherText && otherText.length > 0) {


554                segments.push({ type: 'other', text: otherText });


555            }


556        }


557


558        const matchedText = match[0];


559        let segmentType = 'other';


560        let content = '';


561


562        if (match[1]) {


563            // Asterisk content (*action*)


564            segmentType = 'action';


565            content = matchedText.slice(1, -1);


566        } else if (match[2] || match[3] || match[4] || match[5] || match[6] || match[7]) {


567            // Various quote types ("dialogue")


568            segmentType = 'dialogue';


569            content = matchedText.slice(1, -1);


570        }


571


572        // Trim and check for actual content


573        content = content.trim();


574        if (content.length > 0) {


575            segments.push({


576                type: segmentType,


577                text: content,


578            });


579        }


580


581        lastIndex = match.index + matchedText.length;


582    }


583


584    // Add remaining other text after last match


585    if (lastIndex < text.length) {


586        const otherText = text.substring(lastIndex).trim();


587        if (otherText.length > 0) {


588            segments.push({ type: 'other', text: otherText });


589        }


590    }


591


592    // If no segments found and not empty, treat whole text as other text


593    if (segments.length === 0 && text.trim().length > 0) {


594        segments.push({ type: 'other', text: text.trim() });


595    }


596


597    return segments;


598}


599


600async function processTtsQueue() {


601    // Called each moduleWorker iteration to pull chat messages from queue


602    if (currentTtsJob || ttsJobQueue.length <= 0 || audioPaused) {


603        return;


604    }


605


606    console.debug('New message found, running TTS');


607    currentTtsJob = ttsJobQueue.shift();


608


609    // Handle segmented jobs that already have processed text


610    if (currentTtsJob.segmentType && currentTtsJob.segmentText) {


611        const char = currentTtsJob.name;


612        const segmentText = currentTtsJob.segmentText;


613        const segmentType = currentTtsJob.segmentType;


614


615        console.log(`TTS (${segmentType}): ${segmentText}`);


616


617        try {


618            let voiceMapKey = char;


619


620            // If multi-voice is enabled, modify the voice map key based on segment type


621            if (extension_settings.tts.multi_voice_enabled && char !== DEFAULT_VOICE_MARKER) {


622                switch (segmentType) {


623                    case 'dialogue':


624                        voiceMapKey = `${char} ("Quotes")`;


625                        break;


626                    case 'action':


627                        voiceMapKey = `${char} (*Text inside asterisks*)`;


628                        break;


629                    case 'other':


630                    default:


631                        voiceMapKey = `${char} (Other text)`;


632                        break;


633                }


634            }


635


636            const voiceMapEntry = voiceMap[voiceMapKey] === DEFAULT_VOICE_MARKER ? voiceMap[DEFAULT_VOICE_MARKER] : voiceMap[voiceMapKey];


637


638            if (voiceMapEntry === DISABLED_VOICE_MARKER) {


639                const storageKey = `tts_disabled_warned_${char}`;


640                if (!accountStorage.getItem(storageKey) || currentTtsJob.manual) {


641                    accountStorage.setItem(storageKey, 'true');


642                    toastr.info(`TTS voice for ${char} is disabled.`);


643                }


644                currentTtsJob = null;


645                setTimeout(() => wrapper.update(), 0);


646                return;


647            }


648


649            if (!voiceMapEntry) {


650                throw `${char} not in voicemap. Configure character in extension settings voice map`;


651            }


652


653            const voice = await ttsProvider.getVoice(voiceMapEntry);


654            const voiceId = voice.voice_id;


655            if (voiceId == null) {


656                toastr.error(`Specified voice for ${char} was not found. Check the TTS extension settings.`);


657                throw `Unable to attain voiceId for ${char}`;


658            }


659


660            // Pass the full voiceMapKey (e.g., "User ("Quotes")") as well with character name


661            await tts(segmentText, voiceId, char, voiceMapKey);


662        } catch (error) {


663            toastr.error(error.toString());


664            console.error(error);


665            currentTtsJob = null;


666        }


667        return;


668    }


669


670    // Process unsegmented job (first time processing)


671    let text = extension_settings.tts.narrate_translated_only ? (currentTtsJob?.extra?.display_text || currentTtsJob.mes) : currentTtsJob.mes;


672


673    // Substitute macros


674    text = substituteParams(text);


675


676    if (extension_settings.tts.skip_codeblocks) {


677        text = text.replace(/```.*?```/gs, '').trim();


678        text = text.replace(/~~~.*?~~~/gs, '').trim();


679    }


680


681    if (extension_settings.tts.skip_tags) {


682        text = text.replace(/<.*?>[\s\S]*?<\/.*?>/g, '').trim();


683    }


684


685    if (!extension_settings.tts.pass_asterisks) {


686        text = extension_settings.tts.narrate_dialogues_only


687            ? text.replace(/\*[^*]*?(\*|$)/g, '').trim() // remove asterisks content


688            : text.replaceAll('*', '').trim(); // remove just the asterisks


689    }


690


691    if (extension_settings.tts.apply_regex && extension_settings.tts.regex_pattern) {


692        const regex = regexFromString(extension_settings.tts.regex_pattern);


693        if (regex) {


694            // Clean up extra spaces that might be left after removal


695            text = text.replace(regex, '').replace(/\s+/g, ' ').trim();


696        } else {


697            console.warn('Invalid regex pattern:', extension_settings.tts.regex_pattern);


698        }


699    }


700


701    if (extension_settings.tts.narrate_quoted_only) {


702        const partJoiner = (ttsProvider?.separator || ' ... ');


703        text = joinQuotedBlocks(text, { separator: partJoiner, includeQuotes: true });


704    }


705


706    // Remove embedded images


707    text = text.replace(/!\[.*?]\([^)]*\)/g, '');


708


709    if (typeof ttsProvider?.processText === 'function') {


710        text = await ttsProvider.processText(text);


711    }


712


713    // Collapse newlines and spaces into single space


714    text = text.replace(/\s+/g, ' ').trim();


715


716    console.log(`TTS: ${text}`);


717    const char = currentTtsJob.name;


718


719    // Remove character name from start of the line if power user setting is disabled


720    if (char && !power_user.allow_name2_display) {


721        const escapedChar = escapeRegex(char);


722        text = text.replace(new RegExp(`^${escapedChar}:`, 'gm'), '');


723    }


724


725    try {


726        if (!text) {


727            console.warn('Got empty text in TTS queue job.');


728            completeTtsJob();


729            return;


730        }


731


732        // Parse message into segments if multi-voice is enabled


733        const segments = parseMessageSegments(text);


734


735        if (segments.length === 0) {


736            console.warn('No valid segments found in text.');


737            completeTtsJob();


738            return;


739        }


740


741        // Add all segments to the queue as separate jobs (in reverse order so they process in correct order)


742        for (let i = segments.length - 1; i >= 0; i--) {


743            const segmentJob = {


744                name: char,


745                segmentType: segments[i].type,


746                segmentText: segments[i].text,


747                is_user: currentTtsJob.is_user,


748                mes: currentTtsJob.mes,


749                extra: currentTtsJob.extra,


750                id: currentTtsJob.id,


751                manual: currentTtsJob.manual,


752            };


753            ttsJobQueue.unshift(segmentJob);


754        }


755


756        // Clear current job so the segmented jobs can be processed


757        currentTtsJob = null;


758    } catch (error) {


759        toastr.error(error.toString());


760        console.error(error);


761        currentTtsJob = null;


762    }


763}


764


765/**


766 * Extract and join quoted blocks with proper matching pairs and nesting.


767 * - Captures outermost quotes and everything inside (including different inner quote styles).


768 * - Requires matching opener/closer style (e.g., “ ... ”, 「 ... 」, « ... », etc.).


769 * - Ignores incomplete/unclosed quotes (doesn't include them in the result).


770 * - Symmetric quotes like "..." and ＂...＂ are supported (not nesting the same symmetric style).


771 *


772 * @param {string} text - The text to process


773 * @param {object} [opts={}] - Optional options object


774 * @param {string} [opts.separator=' ... '] - String to join multiple quoted blocks


775 * @param {boolean} [opts.includeQuotes=true] - Keep the quote chars around the captured text


776 * @param {boolean} [opts.returnEmptyOnNoQuotes=false] - Return an empty string if no quotes are found


777 * @param {Array<[string,string]>} [opts.pairs] - Custom quote pairs; defaults cover EN/DE/FR/JP


778 * @returns {string} The joined quoted blocks, or the original text if no quotes found


779 */


780function joinQuotedBlocks(text, opts = {}) {


781    const {


782        separator = ' ... ',


783        includeQuotes = true,


784        returnEmptyOnNoQuotes = false,


785        pairs = [


786            // typographic doubles


787            ['„', '“'],          // DE low-high


788            ['“', '”'],          // EN


789            ['«', '»'],          // FR open « close »


790            ['»', '«'],          // Some locales open »


791            // typographic singles


792            ['‘', '’'],


793            ['‚', '‘'],


794            // Japanese corner quotes


795            ['「', '」'],


796            ['『', '』'],


797            // symmetric doubles


798            ['"', '"'],


799            ['＂', '＂'],


800        ],


801    } = opts;


802


803    if (!text || typeof text !== 'string') return text;


804


805    const openToClose = Object.fromEntries(pairs);


806


807    const segments = [];


808    const stack = []; // [{ opener, expectedClose, start }]


809    for (let i = 0; i < text.length; i++) {


810        const ch = text[i];


811        const top = stack[stack.length - 1];


812


813        // Prefer closing the current open pair if the char matches its expected closer


814        if (top && ch === top.expectedClose) {


815            const finished = stack.pop();


816            if (stack.length === 0) {


817                // Only collect outermost quotes (contains all nested content)


818                segments.push(text.slice(finished.start, i + 1));


819            }


820            continue;


821        }


822


823        // Otherwise, see if this is a new opener


824        if (openToClose[ch]) {


825            stack.push({ opener: ch, expectedClose: openToClose[ch], start: i });


826            continue;


827        }


828


829        // If it's a stray closer that doesn't match current top, ignore


830    }


831


832    if (!segments.length) return returnEmptyOnNoQuotes ? '' : text;


833


834    const cleaned = includeQuotes


835        ? segments


836        : segments.map(s => s.slice(1, -1)); // all defined pairs are single-char quotes


837


838    return cleaned.join(separator);


839}


840


841async function playFullConversation() {


842    resetTtsPlayback();


843


844    if (!extension_settings.tts.enabled) {


845        return toastr.warning('TTS is disabled. Please enable it in the extension settings.');


846    }


847


848    const context = getContext();


849


850    context.chat.forEach((msg, i) => {


851        if (!msg.is_system && msg.mes !== '...' && msg.mes !== '') {


852            processAndQueueTtsMessage(msg, i, { manual: false });


853        }


854    });


855


856    if (ttsJobQueue.length === 0) {


857        return toastr.info('No messages to narrate.');


858    }


859}


860


861globalThis.playFullConversation = playFullConversation;


862


863//#############################//


864//  Extension UI and Settings  //


865//#############################//


866


867function loadSettings() {


868    if (Object.keys(extension_settings.tts).length === 0) {


869        Object.assign(extension_settings.tts, defaultSettings);


870    }


871    for (const key in defaultSettings) {


872        if (!(key in extension_settings.tts)) {


873            extension_settings.tts[key] = defaultSettings[key];


874        }


875    }


876    $('#tts_provider').val(extension_settings.tts.currentProvider);


877    $('#tts_enabled').prop(


878        'checked',


879        extension_settings.tts.enabled,


880    );


881    $('#tts_narrate_dialogues').prop('checked', extension_settings.tts.narrate_dialogues_only);


882    $('#tts_narrate_quoted').prop('checked', extension_settings.tts.narrate_quoted_only);


883    $('#tts_auto_generation').prop('checked', extension_settings.tts.auto_generation);


884    $('#tts_periodic_auto_generation').prop('checked', extension_settings.tts.periodic_auto_generation);


885    $('#tts_narrate_by_paragraphs').prop('checked', extension_settings.tts.narrate_by_paragraphs);


886    $('#tts_narrate_translated_only').prop('checked', extension_settings.tts.narrate_translated_only);


887    $('#tts_narrate_user').prop('checked', extension_settings.tts.narrate_user);


888    $('#tts_pass_asterisks').prop('checked', extension_settings.tts.pass_asterisks);


889    $('#tts_skip_codeblocks').prop('checked', extension_settings.tts.skip_codeblocks);


890    $('#tts_skip_tags').prop('checked', extension_settings.tts.skip_tags);


891    $('#tts_multi_voice_enabled').prop('checked', extension_settings.tts.multi_voice_enabled);


892    $('#tts_apply_regex').prop('checked', extension_settings.tts.apply_regex);


893    $('#tts_regex_pattern').val(extension_settings.tts.regex_pattern);


894    $('#tts_regex_block').toggle(extension_settings.tts.apply_regex);


895    updateRegexPatternWarning();


896    $('#playback_rate').val(extension_settings.tts.playback_rate);


897    $('#playback_rate_counter').val(Number(extension_settings.tts.playback_rate).toFixed(2));


898    $('#playback_rate_block').toggle(extension_settings.tts.currentProvider !== 'System');


899


900    $('body').toggleClass('tts', extension_settings.tts.enabled);


901}


902


903const defaultSettings = {


904    voiceMap: '',


905    ttsEnabled: false,


906    currentProvider: 'ElevenLabs',


907    auto_generation: true,


908    narrate_user: false,


909    playback_rate: 1,


910    multi_voice_enabled: false,


911    apply_regex: false,


912    regex_pattern: '',


913};


914


915function setTtsStatus(status, success) {


916    $('#tts_status').text(status);


917    if (success) {


918        $('#tts_status').removeAttr('style');


919    } else {


920        $('#tts_status').css('color', 'red');


921    }


922}


923


924function onRefreshClick() {


925    Promise.all([


926        ttsProvider.onRefreshClick(),


927        // updateVoiceMap()


928    ]).then(() => {


929        extension_settings.tts[ttsProviderName] = ttsProvider.settings;


930        saveSettingsDebounced();


931        setTtsStatus('Successfully applied settings', true);


932        console.info(`Saved settings ${ttsProviderName} ${JSON.stringify(ttsProvider.settings)}`);


933        initVoiceMap();


934        updateVoiceMap();


935    }).catch(error => {


936        toastr.error(error.toString());


937        console.error(error);


938        setTtsStatus(error, false);


939    });


940}


941


942function onEnableClick() {


943    extension_settings.tts.enabled = $('#tts_enabled').is(


944        ':checked',


945    );


946    updateUiAudioPlayState();


947    saveSettingsDebounced();


948    $('body').toggleClass('tts', extension_settings.tts.enabled);


949}


950


951


952function onAutoGenerationClick() {


953    extension_settings.tts.auto_generation = !!$('#tts_auto_generation').prop('checked');


954    saveSettingsDebounced();


955}


956


957


958function onPeriodicAutoGenerationClick() {


959    extension_settings.tts.periodic_auto_generation = !!$('#tts_periodic_auto_generation').prop('checked');


960    saveSettingsDebounced();


961}


962


963function onNarrateByParagraphsClick() {


964    extension_settings.tts.narrate_by_paragraphs = !!$('#tts_narrate_by_paragraphs').prop('checked');


965    saveSettingsDebounced();


966}


967


968


969function onNarrateDialoguesClick() {


970    extension_settings.tts.narrate_dialogues_only = !!$('#tts_narrate_dialogues').prop('checked');


971    saveSettingsDebounced();


972}


973


974function onNarrateUserClick() {


975    extension_settings.tts.narrate_user = !!$('#tts_narrate_user').prop('checked');


976    saveSettingsDebounced();


977}


978


979function onNarrateQuotedClick() {


980    extension_settings.tts.narrate_quoted_only = !!$('#tts_narrate_quoted').prop('checked');


981    saveSettingsDebounced();


982}


983


984


985function onNarrateTranslatedOnlyClick() {


986    extension_settings.tts.narrate_translated_only = !!$('#tts_narrate_translated_only').prop('checked');


987    saveSettingsDebounced();


988}


989


990function onSkipCodeblocksClick() {


991    extension_settings.tts.skip_codeblocks = !!$('#tts_skip_codeblocks').prop('checked');


992    saveSettingsDebounced();


993}


994


995function onSkipTagsClick() {


996    extension_settings.tts.skip_tags = !!$('#tts_skip_tags').prop('checked');


997    saveSettingsDebounced();


998}


999


1000function onPassAsterisksClick() {


1001    extension_settings.tts.pass_asterisks = !!$('#tts_pass_asterisks').prop('checked');


1002    saveSettingsDebounced();


1003    console.log('setting pass asterisks', extension_settings.tts.pass_asterisks);


1004}


1005


1006function onMultiVoiceClick() {


1007    extension_settings.tts.multi_voice_enabled = !!$('#tts_multi_voice_enabled').prop('checked');


1008    saveSettingsDebounced();


1009    // Reinitialize voice map to show/hide voices


1010    initVoiceMap();


1011}


1012


1013function onApplyRegexChange() {


1014    extension_settings.tts.apply_regex = !!$('#tts_apply_regex').prop('checked');


1015    saveSettingsDebounced();


1016    $('#tts_regex_block').toggle(extension_settings.tts.apply_regex);


1017    updateRegexPatternWarning();


1018}


1019


1020function onRegexPatternChange() {


1021    extension_settings.tts.regex_pattern = $('#tts_regex_pattern').val().toString();


1022    saveSettingsDebounced();


1023    updateRegexPatternWarning();


1024}


1025


1026function updateRegexPatternWarning() {


1027    const warning = $('#tts_regex_warning');


1028    if (!extension_settings.tts.apply_regex) {


1029        warning.hide();


1030        return;


1031    }


1032


1033    const pattern = extension_settings.tts.regex_pattern;


1034    if (!pattern) {


1035        warning.hide();


1036        return;


1037    }


1038


1039    const regex = regexFromString(pattern);


1040    warning.toggle(!regex);


1041}


1042


1043//##############//


1044// TTS Provider //


1045//##############//


1046


1047async function loadTtsProvider(provider) {


1048    //Clear the current config and add new config


1049    $('#tts_provider_settings').html('');


1050


1051    if (!provider) {


1052        return;


1053    }


1054


1055    // Init provider references


1056    extension_settings.tts.currentProvider = provider;


1057    ttsProviderName = provider;


1058    ttsProvider = new ttsProviders[provider];


1059


1060    // Init provider settings


1061    $('#tts_provider_settings').append(ttsProvider.settingsHtml);


1062    if (!(ttsProviderName in extension_settings.tts)) {


1063        console.warn(`Provider ${ttsProviderName} not in Extension Settings, initiatilizing provider in settings`);


1064        extension_settings.tts[ttsProviderName] = {};


1065    }


1066    await ttsProvider.loadSettings(extension_settings.tts[ttsProviderName]);


1067    await initVoiceMap();


1068}


1069


1070function onTtsProviderChange() {


1071    if (typeof ttsProvider?.dispose === 'function') {


1072        ttsProvider.dispose();


1073    }


1074    const ttsProviderSelection = $('#tts_provider').val();


1075    extension_settings.tts.currentProvider = ttsProviderSelection;


1076    $('#playback_rate_block').toggle(extension_settings.tts.currentProvider !== 'System');


1077    loadTtsProvider(ttsProviderSelection);


1078}


1079


1080// Ensure that TTS provider settings are saved to extension settings.


1081export function saveTtsProviderSettings() {


1082    extension_settings.tts[ttsProviderName] = ttsProvider.settings;


1083    updateVoiceMap();


1084    saveSettingsDebounced();


1085    console.info(`Saved settings ${ttsProviderName} ${JSON.stringify(ttsProvider.settings)}`);


1086}


1087


1088


1089//###################//


1090// voiceMap Handling //


1091//###################//


1092


1093async function onChatChanged() {


1094    await onGenerationEnded();


1095    resetTtsPlayback();


1096    const voiceMapInit = initVoiceMap();


1097    await Promise.race([voiceMapInit, delay(debounce_timeout.relaxed)]);


1098    lastMessage = null;


1099}


1100


1101async function onMessageEvent(messageId, lastCharIndex) {


1102    // If TTS is disabled, do nothing


1103    if (!extension_settings.tts.enabled) {


1104        return;


1105    }


1106


1107    // Auto generation is disabled


1108    if (!extension_settings.tts.auto_generation) {


1109        return;


1110    }


1111


1112    const context = getContext();


1113


1114    // no characters or group selected


1115    if (!context.groupId && context.characterId === undefined) {


1116        return;


1117    }


1118


1119    // Chat changed


1120    if (context.chatId !== lastChatId) {


1121        lastChatId = context.chatId;


1122        lastMessageHash = getStringHash(context.chat[messageId]?.mes ?? '');


1123


1124        // Force to speak on the first message in the new chat


1125        if (context.chat.length === 1) {


1126            lastMessageHash = -1;


1127        }


1128    }


1129


1130    // clone message object, as things go haywire if message object is altered below (it's passed by reference)


1131    /** @type {TtsMessage} */


1132    const message = structuredClone(context.chat[messageId]);


1133    const hashNew = getStringHash(message?.mes ?? '');


1134


1135    // Ignore prompt-hidden messages


1136    if (message.is_system) {


1137        return;


1138    }


1139


1140    // if no new messages, or same message, or same message hash, do nothing


1141    if (hashNew === lastMessageHash) {


1142        return;


1143    }


1144


1145    // if we only want to process part of the message


1146    if (lastCharIndex) {


1147        message.mes = message.mes.substring(0, lastCharIndex);


1148    }


1149


1150    const isLastMessageInCurrent = () =>


1151        lastMessage &&


1152        typeof lastMessage === 'object' &&


1153        message.swipe_id === lastMessage.swipe_id &&


1154        message.name === lastMessage.name &&


1155        message.is_user === lastMessage.is_user &&


1156        message.mes.indexOf(lastMessage.mes) !== -1;


1157


1158    // if last message within current message, message got extended. only send diff to TTS.


1159    if (isLastMessageInCurrent()) {


1160        const tmp = structuredClone(message);


1161        message.mes = message.mes.replace(lastMessage.mes, '');


1162        lastMessage = tmp;


1163    } else {


1164        lastMessage = structuredClone(message);


1165    }


1166


1167    // We're currently swiping. Don't generate voice


1168    if (!message || message.mes === '...' || message.mes === '') {


1169        return;


1170    }


1171


1172    // Don't generate if message doesn't have a display text


1173    if (extension_settings.tts.narrate_translated_only && !(message?.extra?.display_text)) {


1174        return;


1175    }


1176


1177    // Don't generate if message is a user message and user message narration is disabled


1178    if (message.is_user && !extension_settings.tts.narrate_user) {


1179        return;


1180    }


1181


1182    // New messages, add new chat to history


1183    lastMessageHash = hashNew;


1184    lastChatId = context.chatId;


1185


1186    console.debug(`Adding message from ${message.name} for TTS processing: "${message.mes}"`);


1187


1188    if (extension_settings.tts.periodic_auto_generation && isStreamingEnabled()) {


1189        message.id = messageId;


1190        ttsJobQueue.push(message);


1191    } else {


1192        processAndQueueTtsMessage(message, messageId, { manual: false });


1193    }


1194}


1195


1196async function onMessageDeleted() {


1197    const context = getContext();


1198


1199    // update internal references to new last message


1200    lastChatId = context.chatId;


1201


1202    // compare against lastMessageHash. If it's the same, we did not delete the last chat item, so no need to reset tts queue


1203    const messageHash = getStringHash((context.chat.length && context.chat[context.chat.length - 1].mes) ?? '');


1204    if (messageHash === lastMessageHash) {


1205        return;


1206    }


1207    lastMessageHash = messageHash;


1208    lastMessage = context.chat.length ? structuredClone(context.chat[context.chat.length - 1]) : null;


1209


1210    // stop any tts playback since message might not exist anymore


1211    resetTtsPlayback();


1212}


1213


1214async function onGenerationStarted(generationType, _args, isDryRun) {


1215    // If dry running or quiet mode, do nothing


1216    if (isDryRun || ['quiet', 'impersonate'].includes(generationType)) {


1217        return;


1218    }


1219


1220    // If TTS is disabled, do nothing


1221    if (!extension_settings.tts.enabled) {


1222        return;


1223    }


1224


1225    // Auto generation is disabled


1226    if (!extension_settings.tts.auto_generation) {


1227        return;


1228    }


1229


1230    // Periodic auto generation is disabled


1231    if (!extension_settings.tts.periodic_auto_generation) {


1232        return;


1233    }


1234


1235    // If the reply is not being streamed


1236    if (!isStreamingEnabled()) {


1237        return;


1238    }


1239


1240    // start the timer


1241    if (!periodicMessageGenerationTimer) {


1242        periodicMessageGenerationTimer = setInterval(onPeriodicMessageGenerationTick, UPDATE_INTERVAL);


1243    }


1244}


1245


1246async function onGenerationEnded() {


1247    if (periodicMessageGenerationTimer) {


1248        clearInterval(periodicMessageGenerationTimer);


1249        periodicMessageGenerationTimer = null;


1250    }


1251    lastPositionOfParagraphEnd = -1;


1252}


1253


1254async function onPeriodicMessageGenerationTick() {


1255    const context = getContext();


1256


1257    // no characters or group selected


1258    if (!context.groupId && context.characterId === undefined) {


1259        return;


1260    }


1261


1262    const lastMessageId = context.chat.length - 1;


1263


1264    // the last message was from the user


1265    if (context.chat[lastMessageId].is_user) {


1266        return;


1267    }


1268


1269    const lastMessage = structuredClone(context.chat[lastMessageId]);


1270    const lastMessageText = lastMessage?.mes ?? '';


1271


1272    // look for double ending lines which should indicate the end of a paragraph


1273    let newLastPositionOfParagraphEnd = lastMessageText


1274        .indexOf('\n\n', lastPositionOfParagraphEnd + 1);


1275    // if not found, look for a single ending line which should indicate the end of a paragraph


1276    if (newLastPositionOfParagraphEnd === -1) {


1277        newLastPositionOfParagraphEnd = lastMessageText


1278            .indexOf('\n', lastPositionOfParagraphEnd + 1);


1279    }


1280


1281    // send the message to the tts module if we found the new end of a paragraph


1282    if (newLastPositionOfParagraphEnd > -1) {


1283        onMessageEvent(lastMessageId, newLastPositionOfParagraphEnd);


1284


1285        if (periodicMessageGenerationTimer) {


1286            lastPositionOfParagraphEnd = newLastPositionOfParagraphEnd;


1287        }


1288    }


1289}


1290


1291/**


1292 * Get characters in current chat


1293 * @param {boolean} unrestricted - If true, will include all characters in voiceMapEntries, even if they are not in the current chat.


1294 * @returns {string[]} - Array of character names


1295 */


1296export function getCharacters(unrestricted) {


1297    const context = getContext();


1298


1299    if (unrestricted) {


1300        const names = context.characters.map(char => char.name);


1301        names.unshift(DEFAULT_VOICE_MARKER);


1302        return names.filter(onlyUnique);


1303    }


1304


1305    let characters = [];


1306    if (context.groupId === null) {


1307        // Single char chat


1308        characters.push(DEFAULT_VOICE_MARKER);


1309        characters.push(context.name1);


1310        characters.push(context.name2);


1311    } else {


1312        // Group chat


1313        characters.push(DEFAULT_VOICE_MARKER);


1314        characters.push(context.name1);


1315        const group = context.groups.find(group => context.groupId == group.id);


1316        for (let member of group.members) {


1317            const character = context.characters.find(char => char.avatar == member);


1318            if (character) {


1319                characters.push(character.name);


1320            }


1321        }


1322    }


1323    characters = characters.filter(onlyUnique);


1324


1325    // If multi-voice is enabled, expand characters to include segment types


1326    if (extension_settings.tts.multi_voice_enabled) {


1327        const expandedCharacters = [];


1328        for (const char of characters) {


1329            if (char === DEFAULT_VOICE_MARKER || char === 'SillyTavern System') {


1330                expandedCharacters.push(char);


1331            } else {


1332                expandedCharacters.push(`${char} ("Quotes")`);


1333                expandedCharacters.push(`${char} (*Text inside asterisks*)`);


1334                expandedCharacters.push(`${char} (Other text)`);


1335            }


1336        }


1337        return expandedCharacters;


1338    }


1339


1340    return characters;


1341}


1342


1343export function sanitizeId(input) {


1344    // Remove any non-alphanumeric characters except underscore (_) and hyphen (-)


1345    let sanitized = encodeURIComponent(input).replace(/[^a-zA-Z0-9-_]/g, '');


1346


1347    // Ensure first character is always a letter


1348    if (!/^[a-zA-Z]/.test(sanitized)) {


1349        sanitized = 'element_' + sanitized;


1350    }


1351


1352    return sanitized;


1353}


1354


1355function parseVoiceMap(voiceMapString) {


1356    let parsedVoiceMap = {};


1357    for (const [charName, voiceId] of voiceMapString


1358        .split(',')


1359        .map(s => s.split(':'))) {


1360        if (charName && voiceId) {


1361            parsedVoiceMap[charName.trim()] = voiceId.trim();


1362        }


1363    }


1364    return parsedVoiceMap;


1365}


1366


1367


1368/**


1369 * Apply voiceMap based on current voiceMapEntries


1370 */


1371function updateVoiceMap() {


1372    const tempVoiceMap = {};


1373    for (const voice of voiceMapEntries) {


1374        if (voice.voiceId === null) {


1375            continue;


1376        }


1377        tempVoiceMap[voice.name] = voice.voiceId;


1378    }


1379    if (Object.keys(tempVoiceMap).length !== 0) {


1380        voiceMap = tempVoiceMap;


1381        console.log(`Voicemap updated to ${JSON.stringify(voiceMap)}`);


1382    }


1383    if (!extension_settings.tts[ttsProviderName].voiceMap) {


1384        extension_settings.tts[ttsProviderName].voiceMap = {};


1385    }


1386    Object.assign(extension_settings.tts[ttsProviderName].voiceMap, voiceMap);


1387    saveSettingsDebounced();


1388}


1389


1390class VoiceMapEntry {


1391    name;


1392    voiceId;


1393    selectElement;


1394    constructor(name, voiceId = DEFAULT_VOICE_MARKER) {


1395        this.name = name;


1396        this.voiceId = voiceId;


1397        this.selectElement = null;


1398    }


1399


1400    addUI(voiceIds) {


1401        let sanitizedName = sanitizeId(this.name);


1402        let defaultOption = this.name === DEFAULT_VOICE_MARKER ?


1403            `<option>${DISABLED_VOICE_MARKER}</option>` :


1404            `<option>${DEFAULT_VOICE_MARKER}</option><option>${DISABLED_VOICE_MARKER}</option>`;


1405        let template = `


1406            <div class='tts_voicemap_block_char flex-container flexGap5'>


1407                <span id='tts_voicemap_char_${sanitizedName}'>${this.name}</span>


1408                <select id='tts_voicemap_char_${sanitizedName}_voice'>


1409                    ${defaultOption}


1410                </select>


1411            </div>


1412        `;


1413        $('#tts_voicemap_block').append(template);


1414


1415        // Populate voice ID select list


1416        for (const voiceId of voiceIds) {


1417            const option = document.createElement('option');


1418            option.innerText = voiceId.name;


1419            option.value = voiceId.name;


1420            $(`#tts_voicemap_char_${sanitizedName}_voice`).append(option);


1421        }


1422


1423        this.selectElement = $(`#tts_voicemap_char_${sanitizedName}_voice`);


1424        this.selectElement.on('change', args => this.onSelectChange(args));


1425        this.selectElement.val(this.voiceId);


1426    }


1427


1428    onSelectChange(args) {


1429        this.voiceId = this.selectElement.find(':selected').val();


1430        updateVoiceMap();


1431    }


1432}


1433


1434/**


1435 * Init voiceMapEntries for character select list.


1436 * If an initialization is already in progress, it returns the existing Promise instead of starting a new one.


1437 * @param {boolean} unrestricted - If true, will include all characters in voiceMapEntries, even if they are not in the current chat.


1438 * @returns {Promise} A promise that resolves when the initialization is complete.


1439 */


1440export async function initVoiceMap(unrestricted = false) {


1441    // Preventing parallel execution


1442    if (currentInitVoiceMapPromise) {


1443        return currentInitVoiceMapPromise;


1444    }


1445


1446    currentInitVoiceMapPromise = (async () => {


1447        const initialChatId = getCurrentChatId();


1448        try {


1449            await initVoiceMapInternal(unrestricted);


1450        } finally {


1451            currentInitVoiceMapPromise = null;


1452        }


1453        const currentChatId = getCurrentChatId();


1454


1455        if (initialChatId !== currentChatId) {


1456            // Chat changed during initialization, reinitialize


1457            await initVoiceMap(unrestricted);


1458        }


1459    })();


1460


1461    return currentInitVoiceMapPromise;


1462}


1463


1464/**


1465 * Init voiceMapEntries for character select list.


1466 * @param {boolean} unrestricted - If true, will include all characters in voiceMapEntries, even if they are not in the current chat.


1467 */


1468async function initVoiceMapInternal(unrestricted) {


1469    // Gate initialization if not enabled or TTS Provider not ready. Prevents error popups.


1470    const enabled = $('#tts_enabled').is(':checked');


1471    if (!enabled) {


1472        return;


1473    }


1474


1475    // Keep errors inside extension UI rather than toastr. Toastr errors for TTS are annoying.


1476    try {


1477        await ttsProvider.checkReady();


1478    } catch (error) {


1479        const message = `TTS Provider not ready. ${error}`;


1480        setTtsStatus(message, false);


1481        return;


1482    }


1483


1484    setTtsStatus('TTS Provider Loaded', true);


1485


1486    // Clear existing voiceMap state


1487    $('#tts_voicemap_block').empty();


1488    voiceMapEntries = [];


1489


1490    // Get characters in current chat


1491    const characters = getCharacters(unrestricted);


1492


1493    // Get saved voicemap from provider settings, handling new and old representations


1494    let voiceMapFromSettings = {};


1495    if ('voiceMap' in extension_settings.tts[ttsProviderName]) {


1496        // Handle previous representation


1497        if (typeof extension_settings.tts[ttsProviderName].voiceMap === 'string') {


1498            voiceMapFromSettings = parseVoiceMap(extension_settings.tts[ttsProviderName].voiceMap);


1499            // Handle new representation


1500        } else if (typeof extension_settings.tts[ttsProviderName].voiceMap === 'object') {


1501            voiceMapFromSettings = extension_settings.tts[ttsProviderName].voiceMap;


1502        }


1503    }


1504


1505    // Get voiceIds from provider


1506    let voiceIdsFromProvider;


1507    try {


1508        voiceIdsFromProvider = await ttsProvider.fetchTtsVoiceObjects();


1509    } catch {


1510        toastr.error('TTS Provider failed to return voice ids.');


1511    }


1512


1513    // Build UI using VoiceMapEntry objects


1514    for (const character of characters) {


1515        if (character === 'SillyTavern System') {


1516            continue;


1517        }


1518        // Check provider settings for voiceIds


1519        let voiceId;


1520        if (character in voiceMapFromSettings) {


1521            voiceId = voiceMapFromSettings[character];


1522        } else if (character === DEFAULT_VOICE_MARKER) {


1523            voiceId = DISABLED_VOICE_MARKER;


1524        } else {


1525            voiceId = DEFAULT_VOICE_MARKER;


1526        }


1527        const voiceMapEntry = new VoiceMapEntry(character, voiceId);


1528        voiceMapEntry.addUI(voiceIdsFromProvider);


1529        voiceMapEntries.push(voiceMapEntry);


1530    }


1531    updateVoiceMap();


1532}


1533


1534export async function init() {


1535    async function addExtensionControls() {


1536        const settingsHtml = $(await renderExtensionTemplateAsync('tts', 'settings'));


1537        $('#tts_container').append(settingsHtml);


1538        $('#tts_refresh').on('click', onRefreshClick);


1539        $('#tts_enabled').on('click', onEnableClick);


1540        $('#tts_narrate_dialogues').on('click', onNarrateDialoguesClick);


1541        $('#tts_narrate_quoted').on('click', onNarrateQuotedClick);


1542        $('#tts_narrate_translated_only').on('click', onNarrateTranslatedOnlyClick);


1543        $('#tts_skip_codeblocks').on('click', onSkipCodeblocksClick);


1544        $('#tts_skip_tags').on('click', onSkipTagsClick);


1545        $('#tts_pass_asterisks').on('click', onPassAsterisksClick);


1546        $('#tts_auto_generation').on('click', onAutoGenerationClick);


1547        $('#tts_periodic_auto_generation').on('click', onPeriodicAutoGenerationClick);


1548        $('#tts_narrate_by_paragraphs').on('click', onNarrateByParagraphsClick);


1549        $('#tts_narrate_user').on('click', onNarrateUserClick);


1550        $('#tts_multi_voice_enabled').on('click', onMultiVoiceClick);


1551        $('#tts_apply_regex').on('change', onApplyRegexChange);


1552        $('#tts_regex_pattern').on('input', onRegexPatternChange);


1553


1554        $('#playback_rate').on('input', function () {


1555            const value = $(this).val();


1556            const formattedValue = Number(value).toFixed(2);


1557            extension_settings.tts.playback_rate = value;


1558            $('#playback_rate_counter').val(formattedValue);


1559            saveSettingsDebounced();


1560        });


1561


1562        $('#tts_voices').on('click', onTtsVoicesClick);


1563        for (const provider in ttsProviders) {


1564            $('#tts_provider').append($('<option />').val(provider).text(provider));


1565        }


1566        $('#tts_provider').on('change', onTtsProviderChange);


1567        $(document).on('click', '.mes_narrate', onNarrateOneMessage);


1568    }


1569    await addExtensionControls(); // No init dependencies


1570    loadSettings(); // Depends on Extension Controls and loadTtsProvider


1571    loadTtsProvider(extension_settings.tts.currentProvider); // No dependencies


1572    addAudioControl(); // Depends on Extension Controls


1573    setInterval(wrapper.update.bind(wrapper), UPDATE_INTERVAL); // Init depends on all the things


1574    eventSource.on(event_types.MESSAGE_SWIPED, resetTtsPlayback);


1575    eventSource.on(event_types.CHAT_CHANGED, onChatChanged);


1576    eventSource.on(event_types.MESSAGE_DELETED, onMessageDeleted);


1577    eventSource.on(event_types.GROUP_UPDATED, onChatChanged);


1578    eventSource.on(event_types.GENERATION_STARTED, onGenerationStarted);


1579    eventSource.on(event_types.GENERATION_ENDED, onGenerationEnded);


1580    eventSource.makeLast(event_types.CHARACTER_MESSAGE_RENDERED, (messageId) => onMessageEvent(messageId));


1581    eventSource.makeLast(event_types.USER_MESSAGE_RENDERED, (messageId) => onMessageEvent(messageId));


1582    SlashCommandParser.addCommandObject(SlashCommand.fromProps({


1583        name: 'speak',


1584        callback: async (args, value) => {


1585            await onNarrateText(args, value);


1586            return '';


1587        },


1588        aliases: ['narrate', 'tts'],


1589        namedArgumentList: [


1590            SlashCommandNamedArgument.fromProps({


1591                name: 'voice',


1592                description: 'character voice name',


1593                typeList: [ARGUMENT_TYPE.STRING],


1594                isRequired: false,


1595                enumProvider: () => Object.keys(voiceMap).map(voiceName => new SlashCommandEnumValue(voiceName, null, enumTypes.enum, enumIcons.voice)),


1596            }),


1597        ],


1598        unnamedArgumentList: [


1599            new SlashCommandArgument(


1600                'text', [ARGUMENT_TYPE.STRING], true,


1601            ),


1602        ],


1603        helpString: `


1604            <div>


1605                Narrate any text using currently selected character's voice.


1606            </div>


1607            <div>


1608                Use <code>voice="Character Name"</code> argument to set other voice from the voice map.


1609            </div>


1610            <div>


1611                <strong>Example:</strong>


1612                <ul>


1613                    <li>


1614                        <pre><code>/speak voice="Donald Duck" Quack!</code></pre>


1615                    </li>


1616                </ul>


1617            </div>


1618        `,


1619    }));


1620


1621    document.body.appendChild(audioElement);


1622}