google tts support api_mode

support gemini tts

.github/workflows/ci.yml

@@ -11,7 +11,7 @@ jobs:
       - uses: actions/checkout@v4
       - uses: actions/setup-node@v4
         with:
-          node-version: lts/*
+          node-version: '20'
       - run: npm install
       - name: Install Docker Compose
         run: |

lib/synth-audio.js

@@ -204,19 +204,23 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
     const startAt = process.hrtime();
     switch (vendor) {
       case 'google':
-        audioData = await synthGoogle(logger, {credentials, stats, language, voice, gender, key, text});
+        audioData = await synthGoogle(logger, {
+          credentials, stats, language, voice, gender, key, text, model, options, instructions,
+          renderForCaching, disableTtsStreaming, disableTtsCache
+        });
         break;
       case 'aws':
       case 'polly':
         vendorLabel = 'aws';
         audioData = await synthPolly(createHash, retrieveHash, logger,
-          {credentials, stats, language, voice, key, text, engine, renderForCaching, disableTtsStreaming});
+          {credentials, stats, language, voice, key, text, engine, renderForCaching, disableTtsStreaming,
+            disableTtsCache});
         break;
       case 'azure':
       case 'microsoft':
         vendorLabel = 'microsoft';
         audioData = await synthMicrosoft(logger, {credentials, stats, language, voice, key, text, deploymentId,
-          renderForCaching, disableTtsStreaming});
+          renderForCaching, disableTtsStreaming, disableTtsCache});
         break;
       case 'nuance':
         model = model || 'enhanced';
@@ -224,7 +228,7 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
         break;
       case 'nvidia':
         audioData = await synthNvidia(client, logger, {credentials, stats, language, voice, model, key, text,
-          renderForCaching, disableTtsStreaming});
+          renderForCaching, disableTtsStreaming, disableTtsCache});
         break;
       case 'ibm':
         audioData = await synthIbm(logger, {credentials, stats, voice, key, text});
@@ -234,44 +238,50 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
         break;
       case 'elevenlabs':
         audioData = await synthElevenlabs(logger, {
-          credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming});
+          credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming,
+          disableTtsCache});
         break;
       case 'playht':
         audioData = await synthPlayHT(client, logger, {
-          credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming});
+          credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming,
+          disableTtsCache});
         break;
       case 'cartesia':
         audioData = await synthCartesia(logger, {
-          credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming});
+          credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming,
+          disableTtsCache});
         break;
       case 'inworld':
         audioData = await synthInworld(logger, {
-          credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming});
+          credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming,
+          disableTtsCache});
         break;
       case 'rimelabs':
         audioData = await synthRimelabs(logger, {
-          credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming});
+          credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming,
+          disableTtsCache});
         break;
       case 'whisper':
         audioData = await synthWhisper(logger, {
-          credentials, stats, voice, key, text, instructions, renderForCaching, disableTtsStreaming});
+          credentials, stats, voice, key, text, instructions, renderForCaching, disableTtsStreaming,
+          disableTtsCache});
         break;
       case 'verbio':
         audioData = await synthVerbio(client, logger, {
-          credentials, stats, voice, key, text, renderForCaching, disableTtsStreaming});
+          credentials, stats, voice, key, text, renderForCaching, disableTtsStreaming, disableTtsCache});
         if (audioData?.filePath) return audioData;
         break;
       case 'deepgram':
         audioData = await synthDeepgram(logger, {credentials, stats, model, key, text,
-          renderForCaching, disableTtsStreaming});
+          renderForCaching, disableTtsStreaming, disableTtsCache});
         break;
       case 'resemble':
         audioData = await synthResemble(logger, {
-          credentials, stats, voice, key, text, options, renderForCaching, disableTtsStreaming});
+          credentials, stats, voice, key, text, options, renderForCaching, disableTtsStreaming, disableTtsCache});
         break;
       case vendor.startsWith('custom') ? vendor : 'cant_match_value':
         audioData = await synthCustomVendor(logger,
-          {credentials, stats, language, voice, key, text});
+          {credentials, stats, language, voice, key, text, renderForCaching, disableTtsStreaming, disableTtsCache});
         break;
       default:
         assert(`synthAudio: unsupported speech vendor ${vendor}`);
@@ -306,14 +316,14 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
 }
 
 const synthPolly = async(createHash, retrieveHash, logger,
-  {credentials, stats, language, voice, engine, key, text, renderForCaching, disableTtsStreaming}) => {
+  {credentials, stats, language, voice, engine, key, text, renderForCaching, disableTtsStreaming, disableTtsCache}) => {
   const {region, accessKeyId, secretAccessKey, roleArn} = credentials;
   if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
 
     let params = '{';
     params += `language=${language}`;
     params += `,playback_id=${key}`;
-    params += ',write_cache_file=1';
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
     params += ',vendor=aws';
     if (accessKeyId && secretAccessKey) {
       if (accessKeyId) params += `,accessKeyId=${accessKeyId}`;
@@ -402,72 +412,123 @@ const synthPolly = async(createHash, retrieveHash, logger,
   }
 };
 
-const synthGoogle = async(logger, {credentials, stats, language, voice, gender, text}) => {
-  const client = new ttsGoogle.TextToSpeechClient(credentials);
-  // If google custom voice cloning is used.
-  // At this time 31 Oct 2024, google node sdk has not support voice cloning yet.
-  if (typeof voice === 'object' && voice.voice_cloning_key) {
-    try {
-      const accessToken = await client.auth.getAccessToken();
-      const projectId = await client.getProjectId();
 
-      const post = bent('https://texttospeech.googleapis.com', 'POST', 'json', {
-        'Authorization': `Bearer ${accessToken}`,
-        'x-goog-user-project': projectId,
-        'Content-Type': 'application/json; charset=utf-8'
-      });
+const synthGoogle = async(logger, {
+  credentials, stats, language, voice, gender, key, text, model, options, instructions,
+  renderForCaching, disableTtsStreaming, disableTtsCache
+}) => {
+  const isGemini = !!model;
+  const isVoiceCloning = typeof voice === 'object' && voice.voice_cloning_key;
+  // HD voices have pattern like en-US-Chirp3-HD-Charon
+  const isHDVoice = typeof voice === 'string' && voice.includes('-HD-');
 
-      const payload = {
-        input: {
-          text
-        },
-        voice: {
-          language_code: language,
-          voice_clone: {
-            voice_cloning_key: voice.voice_cloning_key
-          }
-        },
-        audioConfig: {
-          // Cloning voice at this time still in v1 beta version, and it support LINEAR16 in Wav format, 24.000Hz
-          audioEncoding: 'LINEAR16',
-          sample_rate_hertz: 24000
-        }
-      };
-
-      const wav = await post('/v1beta1/text:synthesize', payload);
-      return {
-        audioContent: Buffer.from(wav.audioContent, 'base64'),
-        extension: 'wav',
-        sampleRate: 24000
-      };
-    } catch (err) {
-      logger.info({err: await err.text()}, 'synthGoogle returned error');
-      throw err;
+  // Streaming support for Google TTS (Gemini, HD voices, and standard voices)
+  // Voice cloning does not support streaming
+  if (!isVoiceCloning && !JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
+    // Strip SSML tags for Gemini TTS (it doesn't support SSML)
+    let inputText = text;
+    if (isGemini && text.startsWith('<speak>')) {
+      inputText = text.replace(/<[^>]*>/g, '').trim();
+      logger.info('synthGoogle: Gemini TTS does not support SSML, stripped tags from input');
     }
+
+    let params = '{';
+    params += `credentials=${Buffer.from(JSON.stringify(credentials.credentials)).toString('base64')}`;
+    params += `,playback_id=${key}`;
+    params += ',vendor=google';
+    params += `,voice=${voice}`;
+    params += `,language_code=${language || 'en-US'}`;
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
+    // api_mode: tts (standard), live (HD voices), gemini (Gemini TTS)
+    const apiMode = options?.apiMode || (isGemini ? 'gemini' : (isHDVoice ? 'live' : 'tts'));
+    params += `,api_mode=${apiMode}`;
+    if (model) params += `,model_name=${model}`;
+    if (gender) params += `,gender=${gender}`;
+    // comma is used to separate parameters in freeswitch tts module
+    const prompt = options?.prompt || instructions;
+    if (prompt) params += `,prompt=${prompt.replace(/\n/g, ' ').replace(/,/g, ';')}`;
+    params += '}';
+
+    return {
+      filePath: `say:${params}${(isGemini ? inputText : text).replace(/\n/g, ' ')}`,
+      servedFromCache: false,
+      rtt: 0
+    };
   }
 
-  const opts = {
-    voice: {
-      ...(typeof voice === 'string' && {name: voice}),
-      ...(typeof voice === 'object' && {customVoice: voice}),
+  const client = new ttsGoogle.TextToSpeechClient(credentials);
+
+  // Build input based on voice type
+  let input;
+  if (isGemini) {
+    // Gemini TTS does not support SSML - strip tags if present
+    let inputText = text;
+    if (text.startsWith('<speak>')) {
+      inputText = text.replace(/<[^>]*>/g, '').trim();
+      logger.info('synthGoogle: Gemini TTS does not support SSML, stripped tags from input');
+    }
+    // Use instructions as prompt for Gemini TTS style control, options.prompt can override
+    const prompt = options?.prompt || instructions;
+    input = {
+      text: inputText,
+      ...(prompt && { prompt })
+    };
+  } else {
+    input = text.startsWith('<speak>') ? { ssml: text } : { text };
+  }
+
+  // Build voice selection params based on voice type
+  let voiceParams;
+  if (isGemini) {
+    voiceParams = {
+      languageCode: language || 'en-US',
+      name: voice,
+      modelName: model
+    };
+  } else if (isVoiceCloning) {
+    voiceParams = {
+      languageCode: language,
+      voiceClone: {
+        voiceCloningKey: voice.voice_cloning_key
+      }
+    };
+  } else {
+    voiceParams = {
+      ...(typeof voice === 'string' && { name: voice }),
+      ...(typeof voice === 'object' && { customVoice: voice }),
       languageCode: language,
       ssmlGender: gender || 'SSML_VOICE_GENDER_UNSPECIFIED'
-    },
-    audioConfig: {audioEncoding: 'MP3'}
-  };
-  Object.assign(opts, {input: text.startsWith('<speak>') ? {ssml: text} : {text}});
+    };
+  }
+
+  // Build audio config based on voice type
+  let audioConfig;
+  let extension;
+  let sampleRate;
+  if (isVoiceCloning) {
+    audioConfig = { audioEncoding: 'LINEAR16', sampleRateHertz: 24000 };
+    extension = 'wav';
+    sampleRate = 24000;
+  } else {
+    audioConfig = { audioEncoding: 'MP3' };
+    extension = 'mp3';
+    sampleRate = 8000;
+  }
+
+  const opts = { input, voice: voiceParams, audioConfig };
+
   try {
-    const responses = await client.synthesizeSpeech(opts);
+    logger.debug({ opts }, 'synthGoogle: request');
+    const [response] = await client.synthesizeSpeech(opts);
     stats.increment('tts.count', ['vendor:google', 'accepted:yes']);
     client.close();
     return {
-      audioContent: responses[0].audioContent,
-      extension: 'mp3',
-      sampleRate: 8000
+      audioContent: response.audioContent,
+      extension,
+      sampleRate
     };
   } catch (err) {
-    console.error(err);
-    logger.info({err, opts}, 'synthAudio: Error synthesizing speech using google');
+    logger.info({ err, opts }, 'synthAudio: Error synthesizing speech using google');
     stats.increment('tts.count', ['vendor:google', 'accepted:no']);
     client && client.close();
     throw err;
@@ -563,7 +624,8 @@ const synthMicrosoft = async(logger, {
   key,
   text,
   renderForCaching,
-  disableTtsStreaming
+  disableTtsStreaming,
+  disableTtsCache
 }) => {
   try {
     const {api_key: apiKey, region, use_custom_tts, custom_tts_endpoint, custom_tts_endpoint_url} = credentials;
@@ -596,7 +658,7 @@ const synthMicrosoft = async(logger, {
       params += `,language=${language}`;
       params += ',vendor=microsoft';
       params += `,voice=${voice}`;
-      params += ',write_cache_file=1';
+      params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
       if (region) params += `,region=${region}`;
       if (custom_tts_endpoint) params += `,endpointId=${custom_tts_endpoint}`;
       if (custom_tts_endpoint_url) params += `,endpoint=${custom_tts_endpoint_url}`;
@@ -769,7 +831,7 @@ const synthNuance = async(client, logger, {credentials, stats, voice, model, tex
 };
 
 const synthNvidia = async(client, logger, {
-  credentials, stats, language,  voice, model, key, text, renderForCaching, disableTtsStreaming
+  credentials, stats, language,  voice, model, key, text, renderForCaching, disableTtsStreaming, disableTtsCache
 }) => {
   const {riva_server_uri} = credentials;
   if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
@@ -778,7 +840,7 @@ const synthNvidia = async(client, logger, {
     params += `,playback_id=${key}`;
     params += `,voice=${voice}`;
     params += `,language=${language}`;
-    params += ',write_cache_file=1';
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
     params += '}';
 
     return {
@@ -818,9 +880,27 @@ const synthNvidia = async(client, logger, {
 };
 
 
-const synthCustomVendor = async(logger, {credentials, stats, language, voice, text, filePath}) => {
+const synthCustomVendor = async(logger, {credentials, stats, language, voice,
+  text, filePath, renderForCaching, disableTtsStreaming, key, disableTtsCache}) => {
   const {vendor, auth_token, custom_tts_url} = credentials;
 
+  if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
+    let params = '{';
+    params += `auth_token=${auth_token}`;
+    params += `,playback_id=${key}`;
+    params += `,custom_tts_url=${custom_tts_url}`;
+    params += ',vendor=custom';
+    params += `,voice=${voice}`;
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
+    params += '}';
+
+    return {
+      filePath: `say:${params}${text.replace(/\n/g, ' ').replace(/\r/g, ' ')}`,
+      servedFromCache: false,
+      rtt: 0
+    };
+  }
+
   try {
     const post = bent('POST', {
       'Authorization': `Bearer ${auth_token}`,
@@ -849,9 +929,9 @@ const synthCustomVendor = async(logger, {credentials, stats, language, voice, te
 };
 
 const synthElevenlabs = async(logger, {
-  credentials, options, stats, voice, key, text, renderForCaching, disableTtsStreaming
+  credentials, options, stats, voice, key, text, renderForCaching, disableTtsStreaming, disableTtsCache
 }) => {
-  const {api_key, model_id, options: credOpts} = credentials;
+  const {api_key, model_id, api_uri, options: credOpts} = credentials;
   const opts = !!options && Object.keys(options).length !== 0 ? options : JSON.parse(credOpts || '{}');
 
   /* default to using the streaming interface, unless disabled by env var OR we want just a cache file */
@@ -862,8 +942,11 @@ const synthElevenlabs = async(logger, {
     params += ',vendor=elevenlabs';
     params += `,voice=${voice}`;
     params += `,model_id=${model_id}`;
-    params += `,optimize_streaming_latency=${opts.optimize_streaming_latency || 2}`;
-    params += ',write_cache_file=1';
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
+    if (api_uri) params += `,api_uri=${api_uri}`;
+    if (opts.optimize_streaming_latency !== null && opts.optimize_streaming_latency !== undefined) {
+      params += `,optimize_streaming_latency=${opts.optimize_streaming_latency}`;
+    }
     if (opts.voice_settings?.similarity_boost) params += `,similarity_boost=${opts.voice_settings.similarity_boost}`;
     if (opts.voice_settings?.stability) params += `,stability=${opts.voice_settings.stability}`;
     if (opts.voice_settings?.style) params += `,style=${opts.voice_settings.style}`;
@@ -886,7 +969,7 @@ const synthElevenlabs = async(logger, {
   const optimize_streaming_latency = opts.optimize_streaming_latency ?
     `?optimize_streaming_latency=${opts.optimize_streaming_latency}` : '';
   try {
-    const post = bent('https://api.elevenlabs.io', 'POST', 'buffer', {
+    const post = bent(`https://${api_uri}`, 'POST', 'buffer', {
       'xi-api-key': api_key,
       'Accept': 'audio/mpeg',
       'Content-Type': 'application/json'
@@ -913,7 +996,7 @@ const synthElevenlabs = async(logger, {
 };
 
 const synthPlayHT = async(client, logger, {
-  credentials, options, stats, voice, language, key, text, renderForCaching, disableTtsStreaming
+  credentials, options, stats, voice, language, key, text, renderForCaching, disableTtsStreaming, disableTtsCache
 }) => {
   const {api_key, user_id, voice_engine, playht_tts_uri, options: credOpts} = credentials;
   const opts = !!options && Object.keys(options).length !== 0 ? options : JSON.parse(credOpts || '{}');
@@ -955,7 +1038,7 @@ const synthPlayHT = async(client, logger, {
     params += `,voice=${voice}`;
     params += `,voice_engine=${voice_engine}`;
     params += `,synthesize_url=${synthesizeUrl}`;
-    params += ',write_cache_file=1';
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
     params += `,language=${language}`;
     if (opts.quality) params += `,quality=${opts.quality}`;
     if (opts.speed) params += `,speed=${opts.speed}`;
@@ -1008,7 +1091,7 @@ const synthPlayHT = async(client, logger, {
 };
 
 const synthInworld = async(logger, {
-  credentials, options, stats, voice, key, text, renderForCaching, disableTtsStreaming
+  credentials, options, stats, voice, key, text, renderForCaching, disableTtsStreaming, disableTtsCache
 }) => {
   const {api_key, model_id, options: credOpts} = credentials;
   const opts = !!options && Object.keys(options).length !== 0 ? options : JSON.parse(credOpts || '{}');
@@ -1021,7 +1104,7 @@ const synthInworld = async(logger, {
     params += `,model_id=${model_id}`;
     params += ',vendor=inworld';
     params += `,voice=${voice}`;
-    params += ',write_cache_file=1';
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
     if (opts.temperature) params += `,temperature=${opts.temperature}`;
     if (opts.audioConfig?.pitch) params += `,pitch=${opts.pitch}`;
     if (opts.audioConfig?.speakingRate) params += `,speakingRate=${opts.speakingRate}`;
@@ -1073,7 +1156,7 @@ const synthInworld = async(logger, {
 };
 
 const synthRimelabs = async(logger, {
-  credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming
+  credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming, disableTtsCache
 }) => {
   const {api_key, model_id, options: credOpts} = credentials;
   const opts = !!options && Object.keys(options).length !== 0 ? options : JSON.parse(credOpts || '{}');
@@ -1087,7 +1170,7 @@ const synthRimelabs = async(logger, {
     params += ',vendor=rimelabs';
     params += `,language=${language}`;
     params += `,voice=${voice}`;
-    params += ',write_cache_file=1';
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
     if (opts.speedAlpha) params += `,speed_alpha=${opts.speedAlpha}`;
     if (opts.reduceLatency) params += `,reduce_latency=${opts.reduceLatency}`;
     // Arcana model parameters
@@ -1131,7 +1214,7 @@ const synthRimelabs = async(logger, {
   }
 };
 const synthVerbio = async(client, logger, {
-  credentials, stats, voice, key, text, renderForCaching, disableTtsStreaming
+  credentials, stats, voice, key, text, renderForCaching, disableTtsStreaming, disableTtsCache
 }) => {
   //https://doc.speechcenter.verbio.com/#tag/Text-To-Speech-REST-API
   if (text.length > 2000) {
@@ -1144,7 +1227,7 @@ const synthVerbio = async(client, logger, {
     params += `,playback_id=${key}`;
     params += ',vendor=verbio';
     params += `,voice=${voice}`;
-    params += ',write_cache_file=1';
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
     params += '}';
 
     return {
@@ -1179,7 +1262,7 @@ const synthVerbio = async(client, logger, {
 };
 
 const synthWhisper = async(logger, {credentials, stats, voice, key, text, instructions,
-  renderForCaching, disableTtsStreaming}) => {
+  renderForCaching, disableTtsStreaming, disableTtsCache}) => {
   const {api_key, model_id, baseURL, timeout, speed} = credentials;
   /* if the env is set to stream then bag out, unless we are specifically rendering to generate a cache file */
   if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
@@ -1189,7 +1272,7 @@ const synthWhisper = async(logger, {credentials, stats, voice, key, text, instru
     params += `,model_id=${model_id}`;
     params += ',vendor=whisper';
     params += `,voice=${voice}`;
-    params += ',write_cache_file=1';
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
     if (speed) params += `,speed=${speed}`;
     // comma is used to separated parameters in freeswitch tts module
     if (instructions) params += `,instructions=${instructions.replace(/\n/g, ' ').replace(/,/g, ';')}`;
@@ -1227,7 +1310,8 @@ const synthWhisper = async(logger, {credentials, stats, voice, key, text, instru
   }
 };
 
-const synthDeepgram = async(logger, {credentials, stats, model, key, text, renderForCaching, disableTtsStreaming}) => {
+const synthDeepgram = async(logger, {credentials, stats, model, key, text, renderForCaching,
+  disableTtsStreaming, disableTtsCache}) => {
   const {api_key, deepgram_tts_uri} = credentials;
   if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
     let params = '{';
@@ -1235,7 +1319,7 @@ const synthDeepgram = async(logger, {credentials, stats, model, key, text, rende
     params += `,playback_id=${key}`;
     params += ',vendor=deepgram';
     params += `,voice=${model}`;
-    params += ',write_cache_file=1';
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
     if (deepgram_tts_uri) params += `,endpoint=${deepgram_tts_uri}`;
     params += '}';
 
@@ -1268,7 +1352,7 @@ const synthDeepgram = async(logger, {credentials, stats, model, key, text, rende
 };
 
 const synthCartesia = async(logger, {
-  credentials, options, stats, voice, language, key, text, renderForCaching, disableTtsStreaming
+  credentials, options, stats, voice, language, key, text, renderForCaching, disableTtsStreaming, disableTtsCache
 }) => {
   const {api_key, model_id, embedding, options: credOpts} = credentials;
   const opts = !!options && Object.keys(options).length !== 0 ? options : JSON.parse(credOpts || '{}');
@@ -1280,12 +1364,13 @@ const synthCartesia = async(logger, {
     params += `,model_id=${model_id}`;
     params += ',vendor=cartesia';
     params += `,voice=${voice}`;
-    params += ',write_cache_file=1';
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
     params += `,language=${language}`;
     params += `,voice_mode=${embedding ? 'embedding' : 'id'}`;
     if (embedding) params += `,embedding=${embedding}`;
     if (opts.speed) params += `,speed=${opts.speed}`;
     if (opts.emotion) params += `,emotion=${opts.emotion}`;
+    if (opts.volume) params += `,volume=${opts.volume}`;
     params += '}';
 
     return {
@@ -1298,7 +1383,7 @@ const synthCartesia = async(logger, {
   try {
     const client = new CartesiaClient({ apiKey: api_key });
     const sampleRate = 48000;
-    const mp3 = await client.tts.bytes({
+    const mp3Stream = await client.tts.bytes({
       modelId: model_id,
       transcript: text,
       voice: {
@@ -1311,13 +1396,20 @@ const synthCartesia = async(logger, {
             id: voice
           }
         ),
-        ...(opts.speed || opts.emotion && {
+        ...(model_id === 'sonic-2' && (opts.speed || opts.emotion) && {
           experimentalControls: {
             ...(opts.speed !== null && opts.speed !== undefined && {speed: opts.speed}),
-            ...(opts.emotion && {emotion: opts.emotion}),
+            ...(opts.emotion && {emotion: [opts.emotion]}),
           }
-        })
+        }),
       },
+      ...(model_id === 'sonic-3' && (opts.speed || opts.emotion || opts.volume) && {
+        generationConfig: {
+          ...(opts.volume !== null && opts.volume !== undefined && {volume: opts.volume}),
+          ...(opts.speed !== null && opts.speed !== undefined && {speed: opts.speed}),
+          ...(opts.emotion !== null && opts.emotion !== undefined && {emotion: opts.emotion}),
+        }
+      }),
       language: language,
       outputFormat: {
         container: 'mp3',
@@ -1325,8 +1417,16 @@ const synthCartesia = async(logger, {
         sampleRate
       },
     });
+
+    // bytes() returns a ReadableStream - collect all chunks
+    const chunks = [];
+    for await (const chunk of mp3Stream) {
+      chunks.push(chunk);
+    }
+    const audioBuffer = Buffer.concat(chunks);
+
     return {
-      audioContent: Buffer.from(mp3),
+      audioContent: audioBuffer,
       extension: 'mp3',
       sampleRate
     };
@@ -1339,7 +1439,7 @@ const synthCartesia = async(logger, {
 };
 
 const synthResemble = async(logger, {
-  credentials, options, stats, voice, key, text, renderForCaching, disableTtsStreaming
+  credentials, options, stats, voice, key, text, renderForCaching, disableTtsStreaming, disableTtsCache
 }) => {
   const {api_key, resemble_tts_uri, resemble_tts_use_tls} = credentials;
   const {project_uuid, use_hd} = options || {};
@@ -1351,7 +1451,7 @@ const synthResemble = async(logger, {
     params += `,playback_id=${key}`;
     params += ',vendor=resemble';
     params += `,voice=${voice}`;
-    params += ',write_cache_file=1';
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
     if (project_uuid) params += `,project_uuid=${project_uuid}`;
     if (use_hd) params += `,use_hd=${use_hd}`;
     if (resemble_tts_uri) params += `,endpoint=${resemble_tts_uri}`;

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/speech-utils",
-  "version": "0.2.22",
+  "version": "0.2.28",
   "description": "TTS-related speech utilities for jambonz",
   "main": "index.js",
   "author": "Dave Horton",
@@ -29,8 +29,8 @@
     "23": "^0.0.0",
     "@aws-sdk/client-polly": "^3.496.0",
     "@aws-sdk/client-sts": "^3.496.0",
-    "@cartesia/cartesia-js": "^2.1.0",
-    "@google-cloud/text-to-speech": "^5.5.0",
+    "@cartesia/cartesia-js": "^2.2.7",
+    "@google-cloud/text-to-speech": "^6.4.0",
     "@grpc/grpc-js": "^1.9.14",
     "@jambonz/realtimedb-helpers": "^0.8.7",
     "bent": "^7.3.12",

test/synth.js

@@ -41,6 +41,7 @@ test('Google speech synth tests', async(t) => {
       gender: 'FEMALE',
       text: 'This is a test.  This is only a test',
       salt: 'foo.bar',
+      renderForCaching: true,
     });
     t.ok(!opts.servedFromCache, `successfully synthesized google audio to ${opts.filePath}`);
 
@@ -55,6 +56,7 @@ test('Google speech synth tests', async(t) => {
       language: 'en-GB',
       gender: 'FEMALE',
       text: 'This is a test.  This is only a test',
+      renderForCaching: true,
     });
     t.ok(opts.servedFromCache, `successfully retrieved cached google audio from ${opts.filePath}`);
 
@@ -78,6 +80,7 @@ test('Google speech synth tests', async(t) => {
       language: 'en-GB',
       gender: 'FEMALE',
       text: 'This is a test.  This is only a test',
+      renderForCaching: true,
     });
     t.ok(!opts.servedFromCache, `successfully synthesized google audio regardless of current cache to ${opts.filePath}`);
   } catch (err) {
@@ -114,7 +117,8 @@ GCP_CUSTOM_VOICE_FILE nor GCP_CUSTOM_VOICE_JSON_KEY provided, GCP_CUSTOM_VOICE_M
       voice: {
         reportedUsage: 'REALTIME',
         model: process.env.GCP_CUSTOM_VOICE_MODEL
-      }
+      },
+      renderForCaching: true,
     });
     t.ok(!opts.servedFromCache, `successfully synthesized google custom voice audio to ${opts.filePath}`);
   } catch (err) {
@@ -132,7 +136,7 @@ test('Google speech voice cloning synth tests', async(t) => {
     !process.env.GCP_CUSTOM_VOICE_JSON_KEY ||
     !process.env.GCP_VOICE_CLONING_FILE &&
     !process.env.GCP_VOICE_CLONING_JSON_KEY) {
-    t.pass(`skipping google speech synth tests since neither 
+    t.pass(`skipping google speech synth tests since neither
 GCP_CUSTOM_VOICE_FILE nor GCP_CUSTOM_VOICE_JSON_KEY provided,
 GCP_VOICE_CLONING_FILE nor GCP_VOICE_CLONING_JSON_KEY is not provided`);
     return t.end();
@@ -156,7 +160,8 @@ GCP_VOICE_CLONING_FILE nor GCP_VOICE_CLONING_JSON_KEY is not provided`);
       text: 'This is a test. This is only a test. This is a test. This is only a test. This is a test. This is only a test',
       voice: {
         voice_cloning_key
-      }
+      },
+      renderForCaching: true,
     });
     t.ok(!opts.servedFromCache, `successfully synthesized google voice cloning audio to ${opts.filePath}`);
   } catch (err) {
@@ -166,6 +171,374 @@ GCP_VOICE_CLONING_FILE nor GCP_VOICE_CLONING_JSON_KEY is not provided`);
   client.quit();
 });
 
+test('Google Gemini TTS synth tests', async(t) => {
+  const fn = require('..');
+  const {synthAudio, client} = fn(opts, logger);
+
+  if (!process.env.GCP_FILE && !process.env.GCP_JSON_KEY) {
+    t.pass('skipping Google Gemini TTS synth tests since neither GCP_FILE nor GCP_JSON_KEY provided');
+    return t.end();
+  }
+  try {
+    const str = process.env.GCP_JSON_KEY || fs.readFileSync(process.env.GCP_FILE);
+    const creds = JSON.parse(str);
+    const geminiModel = process.env.GCP_GEMINI_TTS_MODEL || 'gemini-2.5-flash-tts';
+
+    // Test Gemini TTS with model and instructions (both required for Gemini)
+    let result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'Kore',
+      model: geminiModel,
+      text: 'Hello, this is a test of Google Gemini text to speech.',
+      instructions: 'Speak clearly and naturally.',
+      renderForCaching: true,
+    });
+    t.ok(!result.servedFromCache, `successfully synthesized Google Gemini TTS audio to ${result.filePath}`);
+    t.ok(result.filePath.endsWith('.mp3'), 'Gemini TTS audio file has correct extension');
+
+    // Test Gemini TTS with different voice and instructions
+    result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'Charon',
+      model: geminiModel,
+      text: 'Welcome to our service. How can I help you today?',
+      instructions: 'Speak in a warm, friendly and professional tone.',
+      renderForCaching: true,
+    });
+    t.ok(!result.servedFromCache, `successfully synthesized Gemini TTS with instructions to ${result.filePath}`);
+
+    // Test cache retrieval
+    result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'Kore',
+      model: geminiModel,
+      text: 'Hello, this is a test of Google Gemini text to speech.',
+      instructions: 'Speak clearly and naturally.',
+      renderForCaching: true,
+    });
+    t.ok(result.servedFromCache, `successfully retrieved Gemini TTS audio from cache ${result.filePath}`);
+
+    // Test SSML stripping (Gemini doesn't support SSML)
+    result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'Leda',
+      model: geminiModel,
+      text: '<speak>This SSML should be stripped for Gemini TTS.</speak>',
+      instructions: 'Speak naturally.',
+      disableTtsCache: true,
+      renderForCaching: true,
+    });
+    t.ok(!result.servedFromCache, `successfully synthesized Gemini TTS with SSML stripped to ${result.filePath}`);
+
+  } catch (err) {
+    console.error(err);
+    t.end(err);
+  }
+  client.quit();
+});
+
+test('Google TTS streaming tests (!JAMBONES_DISABLE_TTS_STREAMING)', async(t) => {
+  // Ensure streaming is enabled (default behavior)
+  delete process.env.JAMBONES_DISABLE_TTS_STREAMING;
+
+  // Clear require cache to reload config with new env var
+  delete require.cache[require.resolve('../lib/config')];
+  delete require.cache[require.resolve('../lib/synth-audio')];
+  delete require.cache[require.resolve('..')];
+
+  const fn = require('..');
+  const {synthAudio, client} = fn(opts, logger);
+
+  if (!process.env.GCP_FILE && !process.env.GCP_JSON_KEY) {
+    t.pass('skipping Google TTS streaming tests since neither GCP_FILE nor GCP_JSON_KEY provided');
+    return t.end();
+  }
+
+  try {
+    const str = process.env.GCP_JSON_KEY || fs.readFileSync(process.env.GCP_FILE);
+    const creds = JSON.parse(str);
+    const geminiModel = process.env.GCP_GEMINI_TTS_MODEL || 'gemini-2.5-flash-tts';
+
+    // Test 1: Standard voice streaming (use_live_api=0)
+    let result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'en-US-Wavenet-D',
+      gender: 'MALE',
+      text: 'This is a test of standard voice streaming.',
+      disableTtsCache: true
+    });
+    t.ok(result.filePath.startsWith('say:'), 'Standard voice returns streaming say: path');
+    t.ok(result.filePath.includes('vendor=google'), 'Standard voice streaming path contains vendor=google');
+    t.ok(result.filePath.includes('api_mode=tts'), 'Standard voice uses api_mode=tts');
+    t.ok(result.filePath.includes('voice=en-US-Wavenet-D'), 'Standard voice streaming path contains voice');
+    // Verify credentials are base64 encoded (no raw JSON braces that would break FreeSWitch parsing)
+    t.ok(result.filePath.includes('credentials='), 'Standard voice streaming path contains credentials');
+    t.ok(!result.filePath.includes('credentials={'), 'Credentials are not raw JSON (base64 encoded)');
+
+    // Test 2: HD voice streaming (api_mode=live)
+    result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'en-US-Chirp3-HD-Charon',
+      text: 'This is a test of HD voice streaming.',
+      disableTtsCache: true
+    });
+    t.ok(result.filePath.startsWith('say:'), 'HD voice returns streaming say: path');
+    t.ok(result.filePath.includes('vendor=google'), 'HD voice streaming path contains vendor=google');
+    t.ok(result.filePath.includes('api_mode=live'), 'HD voice uses api_mode=live');
+    t.ok(result.filePath.includes('voice=en-US-Chirp3-HD-Charon'), 'HD voice streaming path contains voice');
+
+    // Test 3: Gemini TTS streaming (api_mode=gemini)
+    result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'Kore',
+      model: geminiModel,
+      text: 'This is a test of Gemini TTS streaming.',
+      instructions: 'Speak naturally.',
+      disableTtsCache: true
+    });
+    t.ok(result.filePath.startsWith('say:'), 'Gemini TTS returns streaming say: path');
+    t.ok(result.filePath.includes('vendor=google'), 'Gemini TTS streaming path contains vendor=google');
+    t.ok(result.filePath.includes('api_mode=gemini'), 'Gemini TTS uses api_mode=gemini');
+    t.ok(result.filePath.includes(`model_name=${geminiModel}`), 'Gemini TTS streaming path contains model_name');
+    t.ok(result.filePath.includes('prompt=Speak naturally.'), 'Gemini TTS streaming path contains prompt');
+
+    // Test 4: Gemini TTS with SSML stripping in streaming mode
+    result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'Leda',
+      model: geminiModel,
+      text: '<speak>This SSML should be stripped.</speak>',
+      instructions: 'Speak naturally.',
+      disableTtsCache: true
+    });
+    t.ok(result.filePath.startsWith('say:'), 'Gemini TTS with SSML returns streaming say: path');
+    t.ok(!result.filePath.includes('<speak>'), 'SSML tags are stripped from streaming path');
+    t.ok(result.filePath.includes('This SSML should be stripped.'), 'Text content is preserved after SSML stripping');
+
+    // Test 5: Gemini TTS with prompt containing special characters
+    result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'Kore',
+      model: geminiModel,
+      text: 'Testing special characters in prompt.',
+      options: { prompt: 'Speak in a warm, friendly tone' },
+      disableTtsCache: true
+    });
+    t.ok(result.filePath.startsWith('say:'), 'Gemini TTS with special chars returns streaming say: path');
+    // Commas in prompt should be replaced with semicolons
+    t.ok(result.filePath.includes('prompt=Speak in a warm; friendly tone'), 'Commas in prompt are escaped to semicolons');
+
+    // Test 6: options.apiMode override (force live on standard voice)
+    result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'en-US-Wavenet-D',
+      text: 'Testing apiMode option override to live.',
+      options: { apiMode: 'live' },
+      disableTtsCache: true
+    });
+    t.ok(result.filePath.includes('api_mode=live'), 'options.apiMode=live overrides default for standard voice');
+
+    // Test 7: options.apiMode override (force gemini without model)
+    result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'Kore',
+      text: 'Testing apiMode option override to gemini.',
+      options: { apiMode: 'gemini' },
+      disableTtsCache: true
+    });
+    t.ok(result.filePath.includes('api_mode=gemini'), 'options.apiMode=gemini overrides default');
+
+    // Test 8: options.apiMode override (force tts on HD voice)
+    result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'en-US-Chirp3-HD-Charon',
+      text: 'Testing apiMode option override to tts on HD voice.',
+      options: { apiMode: 'tts' },
+      disableTtsCache: true
+    });
+    t.ok(result.filePath.includes('api_mode=tts'), 'options.apiMode=tts overrides HD voice default');
+
+  } catch (err) {
+    console.error(err);
+    t.end(err);
+  }
+  client.quit();
+});
+
+test('Google TTS non-streaming tests (JAMBONES_DISABLE_TTS_STREAMING=true)', async(t) => {
+  // Enable streaming disable flag
+  process.env.JAMBONES_DISABLE_TTS_STREAMING = 'true';
+
+  // Clear require cache to reload config with new env var
+  delete require.cache[require.resolve('../lib/config')];
+  delete require.cache[require.resolve('../lib/synth-audio')];
+  delete require.cache[require.resolve('..')];
+
+  const fn = require('..');
+  const {synthAudio, client} = fn(opts, logger);
+
+  if (!process.env.GCP_FILE && !process.env.GCP_JSON_KEY) {
+    t.pass('skipping Google TTS non-streaming tests since neither GCP_FILE nor GCP_JSON_KEY provided');
+    delete process.env.JAMBONES_DISABLE_TTS_STREAMING;
+    return t.end();
+  }
+
+  try {
+    const str = process.env.GCP_JSON_KEY || fs.readFileSync(process.env.GCP_FILE);
+    const creds = JSON.parse(str);
+    const geminiModel = process.env.GCP_GEMINI_TTS_MODEL || 'gemini-2.5-flash-tts';
+
+    // Test 1: Standard voice falls back to non-streaming API
+    let result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'en-US-Wavenet-D',
+      gender: 'MALE',
+      text: 'This is a test with streaming disabled.',
+      disableTtsCache: true
+    });
+    t.ok(!result.filePath.startsWith('say:'), 'Standard voice does NOT return streaming say: path when disabled');
+    t.ok(result.filePath.endsWith('.mp3'), 'Standard voice returns mp3 file path');
+
+    // Test 2: HD voice falls back to non-streaming API
+    result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'en-US-Chirp3-HD-Charon',
+      text: 'This is a test of HD voice with streaming disabled.',
+      disableTtsCache: true
+    });
+    t.ok(!result.filePath.startsWith('say:'), 'HD voice does NOT return streaming say: path when disabled');
+    t.ok(result.filePath.endsWith('.mp3'), 'HD voice returns mp3 file path');
+
+    // Test 3: Gemini TTS falls back to non-streaming API
+    result = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+        },
+      },
+      language: 'en-US',
+      voice: 'Kore',
+      model: geminiModel,
+      text: 'This is a test of Gemini TTS with streaming disabled.',
+      instructions: 'Speak naturally.',
+      disableTtsCache: true
+    });
+    t.ok(!result.filePath.startsWith('say:'), 'Gemini TTS does NOT return streaming say: path when disabled');
+    t.ok(result.filePath.endsWith('.mp3'), 'Gemini TTS returns mp3 file path');
+
+  } catch (err) {
+    console.error(err);
+    t.end(err);
+  } finally {
+    // Clean up: restore default behavior
+    delete process.env.JAMBONES_DISABLE_TTS_STREAMING;
+    delete require.cache[require.resolve('../lib/config')];
+    delete require.cache[require.resolve('../lib/synth-audio')];
+    delete require.cache[require.resolve('..')];
+  }
+  client.quit();
+});
+
 test('AWS speech synth tests', async(t) => {
   const fn = require('..');
   const {synthAudio, client} = fn(opts, logger);
@@ -554,6 +927,7 @@ test('Custom Vendor speech synth tests', async(t) => {
       language: 'en-US',
       voice: 'English-US.Female-1',
       text: 'This is a test.  This is only a test',
+      renderForCaching: true
     });
     t.ok(!opts.servedFromCache, `successfully synthesized custom vendor audio to ${opts.filePath}`);
     t.ok(opts.filePath.endsWith('wav'), 'audio is cached as wav file');
@@ -575,6 +949,7 @@ test('Custom Vendor speech synth tests', async(t) => {
       language: 'en-US',
       voice: 'English-US.Female-1',
       text: 'This is a test.  This is only a test',
+      renderForCaching: true
     });
     t.ok(opts.servedFromCache, `successfully get custom vendor cached audio to ${opts.filePath}`);
     t.ok(opts.filePath.endsWith('wav'), 'audio is cached as wav file');
@@ -589,6 +964,7 @@ test('Custom Vendor speech synth tests', async(t) => {
       language: 'en-US',
       voice: 'English-US.Female-1',
       text: '<speak>This is a test.  This is only a test</speak>',
+      renderForCaching: true
     });
     t.ok(!opts.servedFromCache, `successfully synthesized Custom Vendor audio to ${opts.filePath}`);
     obj = await getJSON(`http://127.0.0.1:3100/lastRequest/somethingnew2`);
@@ -713,7 +1089,7 @@ test('Cartesia speech synth tests', async(t) => {
       text,
       renderForCaching: true
     });
-    t.ok(!opts.servedFromCache, `successfully playht eleven audio to ${opts.filePath}`);
+    t.ok(!opts.servedFromCache, `successfully cartesia eleven audio to ${opts.filePath}`);
 
   } catch (err) {
     console.error(JSON.stringify(err));