0.2.30

google tts support api_mode

.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/add-file-to-cache.js

@@ -3,8 +3,29 @@ const {noopLogger, makeSynthKey} = require('./utils');
 const {JAMBONES_TTS_CACHE_DURATION_MINS} = require('./config');
 const EXPIRES = JAMBONES_TTS_CACHE_DURATION_MINS;
 
+function getExtensionAndSampleRate(path) {
+  const match = path.match(/\.([^.]*)$/);
+  if (!match) {
+    //default should be wav file.
+    return ['wav', 8000];
+  }
+
+  const extension = match[1];
+  const sampleRateMap = {
+    r8: 8000,
+    r16: 16000,
+    r24: 24000,
+    r44: 44100,
+    r48: 48000,
+    r96: 96000,
+  };
+
+  const sampleRate = sampleRateMap[extension] || 8000;
+  return [extension, sampleRate];
+}
+
 async function addFileToCache(client, logger, path,
-  {account_sid, vendor, language, voice, deploymentId, engine, text}) {
+  {account_sid, vendor, language, voice, deploymentId, engine, model, text, instructions}) {
   let key;
   logger = logger || noopLogger;
 
@@ -15,10 +36,19 @@ async function addFileToCache(client, logger, path,
       language: language || '',
       voice: voice || deploymentId,
       engine,
+      model,
       text,
+      instructions
     });
+    const [extension, sampleRate] = getExtensionAndSampleRate(path);
     const audioBuffer = await fs.readFile(path);
-    await client.setex(key, EXPIRES, audioBuffer.toString('base64'));
+    await client.setex(key, EXPIRES, JSON.stringify(
+      {
+        audioContent: audioBuffer.toString('base64'),
+        extension,
+        sampleRate
+      }
+    ));
   } catch (err) {
     logger.error(err, 'addFileToCache: Error');
     return;

lib/config.js

@@ -2,13 +2,14 @@ const JAMBONES_TTS_TRIM_SILENCE = process.env.JAMBONES_TTS_TRIM_SILENCE;
 const JAMBONES_DISABLE_TTS_STREAMING = process.env.JAMBONES_DISABLE_TTS_STREAMING;
 const JAMBONES_DISABLE_AZURE_TTS_STREAMING = process.env.JAMBONES_DISABLE_AZURE_TTS_STREAMING;
 const JAMBONES_EAGERLY_PRE_CACHE_AUDIO = process.env.JAMBONES_EAGERLY_PRE_CACHE_AUDIO;
+const JAMBONES_AZURE_ENABLE_SSML = process.env.JAMBONES_AZURE_ENABLE_SSML;
 
 const JAMBONES_HTTP_PROXY_IP = process.env.JAMBONES_HTTP_PROXY_IP;
 const JAMBONES_HTTP_PROXY_PORT = process.env.JAMBONES_HTTP_PROXY_PORT;
 const JAMBONES_TTS_CACHE_DURATION_MINS =
   (parseInt(process.env.JAMBONES_TTS_CACHE_DURATION_MINS) || 4 * 60) * 60; // cache tts for 4 hours
 
-const TMP_FOLDER = '/tmp';
+const TMP_FOLDER = process.env.JAMBONES_TMP_FOLDER || '/tmp';
 
 const HTTP_TIMEOUT = 5000;
 
@@ -21,5 +22,6 @@ module.exports = {
   JAMBONES_TTS_CACHE_DURATION_MINS,
   JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
   TMP_FOLDER,
-  HTTP_TIMEOUT
+  HTTP_TIMEOUT,
+  JAMBONES_AZURE_ENABLE_SSML
 };

lib/get-aws-sts-token.js

@@ -7,22 +7,23 @@ const CACHE_EXPIRY = process.env.AWS_STS_SESSION_RESET_EXPIRY || (EXPIRY - 600);
 
 async function getAwsAuthToken(
   logger, createHash, retrieveHash,
-  {accessKeyId, secretAccessKey, region, roleArn}) {
+  {speech_credential_sid, accessKeyId, secretAccessKey, region, roleArn}) {
   logger = logger || noopLogger;
   try {
-    const key = makeAwsKey(roleArn || accessKeyId);
+    // if incase instance profile is used, speech_credential_sid will be used as key to lookup cache
+    const key = makeAwsKey(roleArn || accessKeyId || speech_credential_sid);
     const obj = await retrieveHash(key);
     if (obj) return {...obj, servedFromCache: true};
-
+    /* access token not found in cache, so generate it using STS */
     let data;
+    let expiry = CACHE_EXPIRY;
     if (roleArn) {
       const stsClient = new STSClient({ region });
       const roleToAssume = { RoleArn: roleArn, RoleSessionName: 'Jambonz_Speech', DurationSeconds: EXPIRY};
       const command = new AssumeRoleCommand(roleToAssume);
 
       data = await stsClient.send(command);
-    } else {
-      /* access token not found in cache, so generate it using STS */
+    } else if (accessKeyId) {
       const stsClient = new STSClient({
         region,
         credentials: {
@@ -32,6 +33,26 @@ async function getAwsAuthToken(
       });
       const command = new GetSessionTokenCommand({DurationSeconds: EXPIRY});
       data = await stsClient.send(command);
+    } else {
+      // instance profile is used.
+      const stsClient = new STSClient({ region });
+      const cred = await stsClient.config.credentials();
+      // method in the AWS SDK automatically fetches credentials using the default credential
+      // provider chain. If the credentials come from an instance profile or an environment
+      // variable, their expiration is controlled by AWS and not explicitly by our code.
+      if (cred && cred.expiration) {
+        const currentTime = new Date();
+        const expiryTime = new Date(cred.expiration);
+        const remainingTimeInSeconds = Math.round((expiryTime - currentTime) / 1000);
+        expiry = remainingTimeInSeconds;
+      }
+      data = {
+        Credentials: {
+          AccessKeyId: cred.accessKeyId,
+          SecretAccessKey: cred.secretAccessKey,
+          SessionToken: cred.sessionToken
+        }
+      };
     }
 
     const credentials = {
@@ -40,9 +61,11 @@ async function getAwsAuthToken(
       sessionToken: data.Credentials.SessionToken,
       securityToken: data.Credentials.SessionToken
     };
-
-    createHash(key, credentials, CACHE_EXPIRY)
-      .catch((err) => logger.error(err, `Error saving hash for key ${key}`));
+    // Only cache if expiry is good
+    if (expiry > 0) {
+      createHash(key, credentials, expiry)
+        .catch((err) => logger.error(err, `Error saving hash for key ${key}`));
+    }
 
     return {...credentials, servedFromCache: false};
   } catch (err) {

lib/purge-tts-cache.js

@@ -12,7 +12,7 @@ const debug = require('debug')('jambonz:realtimedb-helpers');
  * @returns {object} result - {error, purgedCount}
  */
 async function purgeTtsCache(client, logger, {all, account_sid, vendor,
-  language, voice, deploymentId, engine, text} = {all: true}) {
+  language, voice, deploymentId, engine, model, text, instructions} = {all: true}) {
   logger = logger || noopLogger;
 
   let purgedCount = 0, error;
@@ -33,7 +33,9 @@ async function purgeTtsCache(client, logger, {all, account_sid, vendor,
         language: language || '',
         voice: voice || deploymentId,
         engine,
+        model,
         text,
+        instructions
       });
       purgedCount = await client.del(key);
       if (purgedCount === 0) error = 'Specified item not found';

lib/utils.js

@@ -6,7 +6,7 @@ const pool = new Pool('https://auth.crt.nuance.com');
 const NUANCE_AUTH_ENDPOINT = 'tts.api.nuance.com:443';
 const grpc = require('@grpc/grpc-js');
 const formurlencoded = require('form-urlencoded');
-const { JAMBONES_DISABLE_TTS_STREAMING, JAMBONES_TTS_TRIM_SILENCE, TMP_FOLDER, HTTP_TIMEOUT } = require('./config');
+const { TMP_FOLDER, HTTP_TIMEOUT } = require('./config');
 
 const debug = require('debug')('jambonz:realtimedb-helpers');
 /**
@@ -17,50 +17,27 @@ const debug = require('debug')('jambonz:realtimedb-helpers');
 //const nuanceClientMap = new Map();
 
 function makeSynthKey({
-  account_sid = '', vendor, language, voice, engine = '', text,
-  renderForCaching = false}) {
+  account_sid = '',
+  vendor,
+  language,
+  voice,
+  engine = '',
+  model = '',
+  text,
+  instructions = '',
+}) {
   const hash = crypto.createHash('sha1');
-  hash.update(`${language}:${vendor}:${voice}:${engine}:${text}`);
+  hash.update(`${language}:${vendor}:${voice}:${engine}:${model}:${text}:${instructions}`);
   const hexHashKey = hash.digest('hex');
   const accountKey = account_sid ? `:${account_sid}` : '';
-  const namespace = vendor.startsWith('custom') ? vendor : getFileExtension({vendor, renderForCaching});
-  const key = `tts${accountKey}:${namespace}:${hexHashKey}`;
+  const key = `tts${accountKey}:${hexHashKey}`;
   return key;
 }
 
-function makeFilePath({vendor, key, salt = '', renderForCaching = false}) {
-  const extension = getFileExtension({vendor, renderForCaching});
+function makeFilePath({key, salt = '', extension}) {
   return `${TMP_FOLDER}/${key.replace('tts:', `tts-${salt}`)}.${extension}`;
 }
 
-function getFileExtension({vendor, renderForCaching = false}) {
-  const mp3Extension = 'mp3';
-  const r8Extension = 'r8';
-
-  switch (vendor) {
-    case 'azure':
-    case 'microsoft':
-      if (!renderForCaching && !JAMBONES_DISABLE_TTS_STREAMING || JAMBONES_TTS_TRIM_SILENCE) {
-        return r8Extension;
-      } else {
-        return mp3Extension;
-      }
-    case 'deepgram':
-    case 'elevenlabs':
-    case 'rimlabs':
-      if (!renderForCaching && !JAMBONES_DISABLE_TTS_STREAMING) {
-        return r8Extension;
-      } else {
-        return mp3Extension;
-      }
-    case 'nuance':
-    case 'nvidia':
-    case 'verbio':
-      return r8Extension;
-    default:
-      return mp3Extension;
-  }
-}
 
 const noopLogger = {
   info: () => {},
@@ -89,6 +66,11 @@ function makeAwsKey(awsAccessKeyId) {
   return `aws:${hash.digest('hex')}`;
 }
 
+function makePlayhtKey(apiKey) {
+  const hash = crypto.createHash('sha1');
+  hash.update(apiKey);
+  return `playht:${hash.digest('hex')}`;
+}
 function makeVerbioKey(client_id) {
   const hash = crypto.createHash('sha1');
   hash.update(client_id);
@@ -162,6 +144,7 @@ module.exports = {
   makeSynthKey,
   makeNuanceKey,
   makeIbmKey,
+  makePlayhtKey,
   makeAwsKey,
   makeVerbioKey,
   getNuanceAccessToken,

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/speech-utils",
-  "version": "0.1.15",
+  "version": "0.2.30",
   "description": "TTS-related speech utilities for jambonz",
   "main": "index.js",
   "author": "Dave Horton",
@@ -26,22 +26,24 @@
   },
   "homepage": "https://github.com/jambonz/speech-utils#readme",
   "dependencies": {
+    "23": "^0.0.0",
     "@aws-sdk/client-polly": "^3.496.0",
     "@aws-sdk/client-sts": "^3.496.0",
-    "@google-cloud/text-to-speech": "^5.0.2",
+    "@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",
     "debug": "^4.3.4",
     "form-urlencoded": "^6.1.4",
     "google-protobuf": "^3.21.2",
-    "ibm-watson": "^8.0.0",
+    "ibm-watson": "^11.0.0",
     "microsoft-cognitiveservices-speech-sdk": "1.38.0",
-    "openai": "^4.25.0",
-    "undici": "^6.4.0"
+    "openai": "^4.98.0",
+    "undici": "^7.5.0"
   },
   "devDependencies": {
-    "config": "^3.3.11",
+    "config": "^4.2.0",
     "eslint": "^9.3.0",
     "eslint-plugin-promise": "^6.2.0",
     "husky": "^9.0.11",

test/synth.js

@@ -5,7 +5,7 @@ const fs = require('fs');
 const {makeSynthKey} = require('../lib/utils');
 const logger = require('pino')();
 const bent = require('bent');
-const getJSON = bent('json')
+const getJSON = bent('json');
 
 process.on('unhandledRejection', (reason, p) => {
   console.log('Unhandled Rejection at: Promise', p, 'reason:', reason);
@@ -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) {
@@ -91,14 +94,17 @@ test('Google speech Custom voice synth tests', async(t) => {
   const fn = require('..');
   const {synthAudio, client} = fn(opts, logger);
 
-  if (!process.env.GCP_CUSTOM_VOICE_FILE && !process.env.GCP_CUSTOM_VOICE_JSON_KEY || !process.env.GCP_CUSTOM_VOICE_MODEL) {
-    t.pass('skipping google speech synth tests since neither GCP_CUSTOM_VOICE_FILE nor GCP_CUSTOM_VOICE_JSON_KEY provided, GCP_CUSTOM_VOICE_MODEL is not provided');
+  if (!process.env.GCP_CUSTOM_VOICE_FILE &&
+    !process.env.GCP_CUSTOM_VOICE_JSON_KEY ||
+    !process.env.GCP_CUSTOM_VOICE_MODEL) {
+    t.pass(`skipping google speech synth tests since neither 
+GCP_CUSTOM_VOICE_FILE nor GCP_CUSTOM_VOICE_JSON_KEY provided, GCP_CUSTOM_VOICE_MODEL is not provided`);
     return t.end();
   }
   try {
     const str = process.env.GCP_CUSTOM_VOICE_JSON_KEY || fs.readFileSync(process.env.GCP_CUSTOM_VOICE_FILE);
     const creds = JSON.parse(str);
-    let opts = await synthAudio(stats, {
+    const opts = await synthAudio(stats, {
       vendor: 'google',
       credentials: {
         credentials: {
@@ -109,9 +115,10 @@ test('Google speech Custom voice synth tests', async(t) => {
       language: 'en-AU',
       text: 'This is a test.  This is only a test',
       voice: {
-        reportedUsage:"REALTIME",
+        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) {
@@ -121,6 +128,417 @@ test('Google speech Custom voice synth tests', async(t) => {
   client.quit();
 });
 
+test('Google speech voice cloning synth tests', async(t) => {
+  const fn = require('..');
+  const {synthAudio, client} = fn(opts, logger);
+
+  if (!process.env.GCP_CUSTOM_VOICE_FILE &&
+    !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
+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();
+  }
+  try {
+    const googleKey = process.env.GCP_CUSTOM_VOICE_JSON_KEY ||
+      fs.readFileSync(process.env.GCP_CUSTOM_VOICE_FILE);
+    const voice_cloning_key = process.env.GCP_VOICE_CLONING_JSON_KEY ||
+      fs.readFileSync(process.env.GCP_VOICE_CLONING_FILE).toString();
+    const creds = JSON.parse(googleKey);
+    const opts = await synthAudio(stats, {
+      vendor: 'google',
+      credentials: {
+        credentials: {
+          client_email: creds.client_email,
+          private_key: creds.private_key,
+          project_id: creds.project_id
+        },
+      },
+      language: 'en-US',
+      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) {
+    console.error(err);
+    t.end(err);
+  }
+  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);
@@ -140,6 +558,7 @@ test('AWS speech synth tests', async(t) => {
       language: 'en-US',
       voice: 'Joey',
       text: 'This is a test.  This is only a test',
+      renderForCaching: true,
     });
     t.ok(!opts.servedFromCache, `successfully synthesized aws audio to ${opts.filePath}`);
 
@@ -153,6 +572,7 @@ test('AWS speech synth tests', async(t) => {
       language: 'en-US',
       voice: 'Joey',
       text: 'This is a test.  This is only a test',
+      renderForCaching: true,
     });
     t.ok(opts.servedFromCache, `successfully retrieved aws audio from cache ${opts.filePath}`);
   } catch (err) {
@@ -507,8 +927,10 @@ 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');
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/somethingnew`);
     t.ok(obj.headers.Authorization == 'Bearer some_jwt_token', 'Custom Vendor Authentication Header is correct');
     t.ok(obj.body.language == 'en-US', 'Custom Vendor Language is correct');
@@ -516,6 +938,22 @@ test('Custom Vendor speech synth tests', async(t) => {
     t.ok(obj.body.type == 'text', 'Custom Vendor type is correct');
     t.ok(obj.body.text == 'This is a test.  This is only a test', 'Custom Vendor text is correct');
 
+    // Checking if cache is stored with wav format
+    opts = await synthAudio(stats, {
+      vendor: 'custom:somethingnew',
+      credentials: {
+        use_for_tts: 1,
+        custom_tts_url: "http://127.0.0.1:3100/somethingnew",
+        auth_token: 'some_jwt_token'
+      },
+      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');
+
     opts = await synthAudio(stats, {
       vendor: 'custom:somethingnew2',
       credentials: {
@@ -526,6 +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`);
@@ -574,9 +1013,9 @@ test('Elevenlabs speech synth tests', async(t) => {
     t.end(err);
   }
   client.quit();
-})
+});
 
-test('PlayHT speech synth tests', async(t) => {
+const testPlayHT = async(t, voice_engine) => {
   const fn = require('..');
   const {synthAudio, client} = fn(opts, logger);
 
@@ -584,26 +1023,26 @@ test('PlayHT speech synth tests', async(t) => {
     t.pass('skipping PlayHT speech synth tests since PLAYHT_API_KEY or PLAYHT_USER_ID is/are not provided');
     return t.end();
   }
-  const text = 'Hi there and welcome to jambones!';
+  const text = 'Hi there and welcome to jambones! ' + Date.now();
   try {
-    let opts = await synthAudio(stats, {
+    const opts = await synthAudio(stats, {
       vendor: 'playht',
       credentials: {
         api_key: process.env.PLAYHT_API_KEY,
         user_id: process.env.PLAYHT_USER_ID,
-        voice_engine: 'PlayHT2.0-turbo',
+        voice_engine,
         options: JSON.stringify({
-          quality: "medium",
+          quality: 'medium',
           speed: 1,
           seed: 1,
           temperature: 1,
-          emotion: "female_happy",
+          emotion: 'female_happy',
           voice_guidance: 3,
           style_guidance: 20,
           text_guidance: 1,
         })
       },
-      language: 'en-US',
+      language: 'english',
       voice: 's3://voice-cloning-zero-shot/d9ff78ba-d016-47f6-b0ef-dd630f59414e/female-cs/manifest.json',
       text,
       renderForCaching: true
@@ -615,9 +1054,110 @@ test('PlayHT speech synth tests', async(t) => {
     t.end(err);
   }
   client.quit();
+};
+
+test('PlayHT speech synth tests', async(t) => {
+  await testPlayHT(t, 'PlayHT2.0-turbo');
 });
 
-test('rimelabs speech synth tests', async(t) => {
+test('PlayHT3.0 speech synth tests', async(t) => {
+  await testPlayHT(t, 'Play3.0');
+});
+
+test('Cartesia speech synth tests', async(t) => {
+  const fn = require('..');
+  const {synthAudio, client} = fn(opts, logger);
+
+  if (!process.env.CARTESIA_API_KEY) {
+    t.pass('skipping Cartesia speech synth tests since CARTESIA_API_KEY is not provided');
+    return t.end();
+  }
+  const text = 'Hi there and welcome to jambones! ' + Date.now();
+  try {
+    const opts = await synthAudio(stats, {
+      vendor: 'cartesia',
+      credentials: {
+        api_key: process.env.CARTESIA_API_KEY,
+        model_id: 'sonic-english',
+        options: JSON.stringify({
+          speed: 1,
+          emotion: 'female_happy',
+        })
+      },
+      language: 'en',
+      voice: '694f9389-aac1-45b6-b726-9d9369183238',
+      text,
+      renderForCaching: true
+    });
+    t.ok(!opts.servedFromCache, `successfully cartesia eleven audio to ${opts.filePath}`);
+
+  } catch (err) {
+    console.error(JSON.stringify(err));
+    t.end(err);
+  }
+  client.quit();
+});
+
+test('inworld speech synth', async(t) => {
+  const fn = require('..');
+  const {synthAudio, client} = fn(opts, logger);
+
+  if (!process.env.INWORLD_API_KEY) {
+    t.pass('skipping inworld speech synth tests since INWORLD_API_KEY is not provided');
+    return t.end();
+  }
+  const text = 'Hi there and welcome to jambones!';
+  try {
+    const opts = await synthAudio(stats, {
+      vendor: 'inworld',
+      credentials: {
+        api_key: process.env.INWORLD_API_KEY,
+        model_id: 'inworld-tts-1'
+      },
+      language: 'en',
+      voice: 'Ashley',
+      text,
+      renderForCaching: true
+    });
+    t.ok(!opts.servedFromCache, `successfully synthesized inworld audio to ${opts.filePath}`);
+
+  } catch (err) {
+    console.error(JSON.stringify(err));
+    t.end(err);
+  }
+  client.quit();
+});
+
+test('resemble speech synth', async(t) => {
+  const fn = require('..');
+  const {synthAudio, client} = fn(opts, logger);
+
+  if (!process.env.RESEMBLE_API_KEY) {
+    t.pass('skipping resemble speech synth tests since RESEMBLE_API_KEY is not provided');
+    return t.end();
+  }
+  const text = '<speak prompt="Speak in an excited, upbeat tone">Hello from Resemble!</speak>';
+  try {
+    const opts = await synthAudio(stats, {
+      vendor: 'resemble',
+      credentials: {
+        api_key: process.env.RESEMBLE_API_KEY,
+      },
+      language: 'en',
+      voice: '3f5fb9f1',
+      text,
+      renderForCaching: true
+    });
+    t.ok(!opts.servedFromCache, `successfully synthesized resemble audio to ${opts.filePath}`);
+
+  } catch (err) {
+    console.error(JSON.stringify(err));
+    t.end(err);
+  }
+  client.quit();
+});
+
+test('rimelabs speech synth tests mist', async(t) => {
   const fn = require('..');
   const {synthAudio, client} = fn(opts, logger);
 
@@ -627,7 +1167,7 @@ test('rimelabs speech synth tests', async(t) => {
   }
   const text = 'Hi there and welcome to jambones!';
   try {
-    let opts = await synthAudio(stats, {
+    const opts = await synthAudio(stats, {
       vendor: 'rimelabs',
       credentials: {
         api_key: process.env.RIMELABS_API_KEY,
@@ -637,7 +1177,7 @@ test('rimelabs speech synth tests', async(t) => {
           reduceLatency: false
         })
       },
-      language: 'en-US',
+      language: 'eng',
       voice: 'amber',
       text,
       renderForCaching: true
@@ -651,6 +1191,40 @@ test('rimelabs speech synth tests', async(t) => {
   client.quit();
 });
 
+test('rimelabs speech synth tests mistv2', async(t) => {
+  const fn = require('..');
+  const {synthAudio, client} = fn(opts, logger);
+
+  if (!process.env.RIMELABS_API_KEY) {
+    t.pass('skipping rimelabs speech synth tests since RIMELABS_API_KEY is not provided');
+    return t.end();
+  }
+  const text = 'Hi there and welcome to jambones!';
+  try {
+    const opts = await synthAudio(stats, {
+      vendor: 'rimelabs',
+      credentials: {
+        api_key: process.env.RIMELABS_API_KEY,
+        model_id: 'mistv2',
+        options: JSON.stringify({
+          speedAlpha: 1.0,
+          reduceLatency: false
+        })
+      },
+      language: 'spa',
+      voice: 'pablo',
+      text,
+      renderForCaching: true
+    });
+    t.ok(!opts.servedFromCache, `successfully synthesized rimelabs mistv2 audio to ${opts.filePath}`);
+
+  } catch (err) {
+    console.error(JSON.stringify(err));
+    t.end(err);
+  }
+  client.quit();
+});
+
 test('whisper speech synth tests', async(t) => {
   const fn = require('..');
   const {synthAudio, client} = fn(opts, logger);
@@ -750,11 +1324,12 @@ test('TTS Cache tests', async(t) => {
     // save some random tts keys to cache
     const minRecords = 8;
     for (const i in Array(minRecords).fill(0)) {
-      await client.set(makeSynthKey({vendor: i, language: i, voice: i, engine: i, text: i}), i);
+      await client.set(makeSynthKey({vendor: i, language: i, voice: i, engine: i, model: i, text: i,
+        instructions: i}), i);
     }
     const count = await getTtsSize();
     t.ok(count >= minRecords, 'getTtsSize worked.');
-    
+
     const {purgedCount} = await purgeTtsCache();
     t.ok(purgedCount >= minRecords, `successfully purged at least ${minRecords} tts records from cache`);
 
@@ -769,7 +1344,7 @@ test('TTS Cache tests', async(t) => {
   try {
     // save some random tts keys to cache
     for (const i in Array(10).fill(0)) {
-      await client.set(makeSynthKey({vendor: i, language: i, voice: i, engine: i, text: i}), i);
+      await client.set(makeSynthKey({vendor: i, language: i, voice: i, engine: i, text: i, instructions: i}), i);
     }
     // save a specific key to tts cache
     const opts = {vendor: 'aws', language: 'en-US', voice: 'MALE', engine: 'Engine', text: 'Hello World!'};
@@ -785,24 +1360,15 @@ test('TTS Cache tests', async(t) => {
       language: 'non-existing',
       voice: 'non-existing',
     });
-    t.ok(purgedCountWhenErrored === 0, `purged no records when specified key was not found`);
-    t.ok(error, `error returned when specified key was not found`);
+    t.ok(purgedCountWhenErrored === 0, 'purged no records when specified key was not found');
+    t.ok(error, 'error returned when specified key was not found');
 
     // make sure other tts keys are still there
-    const cached = await client.keys('tts:*')
-    t.ok(cached.length >= 1, `successfully kept all non-specified tts records in cache`);
+    const cached = await client.keys('tts:*');
+    t.ok(cached.length >= 1, 'successfully kept all non-specified tts records in cache');
 
-    // retrieve keys from cache and check the key contains the file extension    
-    let key = cached[0];
-    t.ok(key.includes('mp3'), `tts cache extension shoult be part of the key and equal mp3`);
-
-    process.env.VG_TRIM_TTS_SILENCE = 'true';    
+    process.env.VG_TRIM_TTS_SILENCE = 'true';
     await client.set(makeSynthKey({ vendor: 'azure' }), 'value');
-
-    const r8Keys = await client.keys('tts:r8*');
-    key = r8Keys[0];
-    t.ok(key.includes('r8'), `tts cache extension shoult be part of the key and equal r8`);
-
   } catch (err) {
     console.error(JSON.stringify(err));
     t.end(err);
@@ -816,10 +1382,10 @@ test('TTS Cache tests', async(t) => {
     const account_sid = "12412512_cabc_5aff"
     const account_sid2 = "22412512_cabc_5aff"
     for (const i in Array(minRecords).fill(0)) {
-      await client.set(makeSynthKey({account_sid, vendor: i, language: i, voice: i, engine: i, text: i}), i);
+      await client.set(makeSynthKey({account_sid, vendor: i, language: i, voice: i, engine: i, text: i, instructions: i}), i);
     }
     for (const i in Array(minRecords).fill(0)) {
-      await client.set(makeSynthKey({account_sid: account_sid2, vendor: i, language: i, voice: i, engine: i, text: i}), i);
+      await client.set(makeSynthKey({account_sid: account_sid2, vendor: i, language: i, voice: i, engine: i, text: i, instructions: i}), i);
     }
     const {purgedCount} = await purgeTtsCache({account_sid});
     t.equal(purgedCount, minRecords, `successfully purged at least ${minRecords} tts records from cache for account_sid:${account_sid}`);