0.2.4

update undici

.eslintignore

@@ -1 +0,0 @@
-test/*

.eslintrc.json

@@ -1,126 +0,0 @@
-{
-  "env": {
-    "node": true,
-    "es6": true  
-  },
-  "parserOptions": {
-    "ecmaFeatures": {
-      "jsx": false,
-      "modules": false
-    },
-    "ecmaVersion": 2020
-  },
-  "plugins": ["promise"],
-  "rules": {
-    "promise/always-return": "error",
-    "promise/no-return-wrap": "error",
-    "promise/param-names": "error",
-    "promise/catch-or-return": "error",
-    "promise/no-native": "off",
-    "promise/no-nesting": "warn",
-    "promise/no-promise-in-callback": "warn",
-    "promise/no-callback-in-promise": "warn",
-    "promise/no-return-in-finally": "warn",
-
-    // Possible Errors
-    // http://eslint.org/docs/rules/#possible-errors
-    "comma-dangle": [2, "only-multiline"],
-    "no-control-regex": 2,
-    "no-debugger": 2,
-    "no-dupe-args": 2,
-    "no-dupe-keys": 2,
-    "no-duplicate-case": 2,
-    "no-empty-character-class": 2,
-    "no-ex-assign": 2,
-    "no-extra-boolean-cast" : 2,
-    "no-extra-parens": [2, "functions"],
-    "no-extra-semi": 2,
-    "no-func-assign": 2,
-    "no-invalid-regexp": 2,
-    "no-irregular-whitespace": 2,
-    "no-negated-in-lhs": 2,
-    "no-obj-calls": 2,
-    "no-proto": 2,
-    "no-unexpected-multiline": 2,
-    "no-unreachable": 2,
-    "use-isnan": 2,
-    "valid-typeof": 2,
-
-    // Best Practices
-    // http://eslint.org/docs/rules/#best-practices
-    "no-fallthrough": 2,
-    "no-octal": 2,
-    "no-redeclare": 2,
-    "no-self-assign": 2,
-    "no-unused-labels": 2,
-
-    // Strict Mode
-    // http://eslint.org/docs/rules/#strict-mode
-    "strict": [2, "never"],
-
-    // Variables
-    // http://eslint.org/docs/rules/#variables
-    "no-delete-var": 2,
-    "no-undef": 2,
-    "no-unused-vars": [2, {"args": "none"}],
-
-    // Node.js and CommonJS
-    // http://eslint.org/docs/rules/#nodejs-and-commonjs
-    "no-mixed-requires": 2,
-    "no-new-require": 2,
-    "no-path-concat": 2,
-    "no-restricted-modules": [2, "sys", "_linklist"],
-
-    // Stylistic Issues
-    // http://eslint.org/docs/rules/#stylistic-issues
-    "comma-spacing": 2,
-    "eol-last": 2,
-    "indent": [2, 2, {"SwitchCase": 1}],
-    "keyword-spacing": 2,
-    "max-len": [2, 120, 2],
-    "new-parens": 2,
-    "no-mixed-spaces-and-tabs": 2,
-    "no-multiple-empty-lines": [2, {"max": 2}],
-    "no-trailing-spaces": [2, {"skipBlankLines": false }],
-    "quotes": [2, "single", "avoid-escape"],
-    "semi": 2,
-    "space-before-blocks": [2, "always"],
-    "space-before-function-paren": [2, "never"],
-    "space-in-parens": [2, "never"],
-    "space-infix-ops": 2,
-    "space-unary-ops": 2,
-
-    // ECMAScript 6
-    // http://eslint.org/docs/rules/#ecmascript-6
-    "arrow-parens": [2, "always"],
-    "arrow-spacing": [2, {"before": true, "after": true}],
-    "constructor-super": 2,
-    "no-class-assign": 2,
-    "no-confusing-arrow": 2,
-    "no-const-assign": 2,
-    "no-dupe-class-members": 2,
-    "no-new-symbol": 2,
-    "no-this-before-super": 2,
-    "prefer-const": 2
-  },
-  "globals": {
-    "DTRACE_HTTP_CLIENT_REQUEST"           : false,
-    "LTTNG_HTTP_CLIENT_REQUEST"            : false,
-    "COUNTER_HTTP_CLIENT_REQUEST"          : false,
-    "DTRACE_HTTP_CLIENT_RESPONSE"          : false,
-    "LTTNG_HTTP_CLIENT_RESPONSE"           : false,
-    "COUNTER_HTTP_CLIENT_RESPONSE"         : false,
-    "DTRACE_HTTP_SERVER_REQUEST"           : false,
-    "LTTNG_HTTP_SERVER_REQUEST"            : false,
-    "COUNTER_HTTP_SERVER_REQUEST"          : false,
-    "DTRACE_HTTP_SERVER_RESPONSE"          : false,
-    "LTTNG_HTTP_SERVER_RESPONSE"           : false,
-    "COUNTER_HTTP_SERVER_RESPONSE"         : false,
-    "DTRACE_NET_STREAM_END"                : false,
-    "LTTNG_NET_STREAM_END"                 : false,
-    "COUNTER_NET_SERVER_CONNECTION_CLOSE"  : false,
-    "DTRACE_NET_SERVER_CONNECTION"         : false,
-    "LTTNG_NET_SERVER_CONNECTION"          : false,
-    "COUNTER_NET_SERVER_CONNECTION"        : false  
-  }
-}

.github/workflows/ci.yml

@@ -13,6 +13,11 @@ jobs:
         with:
           node-version: lts/*
       - run: npm install
+      - name: Install Docker Compose
+        run: |
+          sudo curl -L "https://github.com/docker/compose/releases/download/1.29.2/docker-compose-$(uname -s)-$(uname -m)" -o /usr/local/bin/docker-compose
+          sudo chmod +x /usr/local/bin/docker-compose
+          docker-compose --version
       - run: npm run jslint
       - run: sudo apt update && sudo apt install -y squid
       - run: sudo cp test/squid.conf /etc/squid/squid.conf

.husky/pre-commit

@@ -0,0 +1,3 @@
+#npm audit
+npm run jslint:fix || true
+npm test

eslint.config.js

@@ -0,0 +1,140 @@
+const promise = require('eslint-plugin-promise');
+const globals = require('globals');
+
+
+module.exports = [
+  {
+    files: ['**/*.js', '*.js'],
+    languageOptions: {
+
+      ecmaVersion: 2020,
+      parserOptions: {
+        'node': true,
+        'es6': true,
+        ecmaFeatures: {
+          jsx: false,
+          modules: false,
+        }
+      },
+      sourceType: 'module',
+      globals: {
+        ...globals.node,
+        'DTRACE_HTTP_CLIENT_REQUEST': false,
+        'LTTNG_HTTP_CLIENT_REQUEST': false,
+        'COUNTER_HTTP_CLIENT_REQUEST': false,
+        'DTRACE_HTTP_CLIENT_RESPONSE': false,
+        'LTTNG_HTTP_CLIENT_RESPONSE': false,
+        'COUNTER_HTTP_CLIENT_RESPONSE': false,
+        'DTRACE_HTTP_SERVER_REQUEST': false,
+        'LTTNG_HTTP_SERVER_REQUEST': false,
+        'COUNTER_HTTP_SERVER_REQUEST': false,
+        'DTRACE_HTTP_SERVER_RESPONSE': false,
+        'LTTNG_HTTP_SERVER_RESPONSE': false,
+        'COUNTER_HTTP_SERVER_RESPONSE': false,
+        'DTRACE_NET_STREAM_END': false,
+        'LTTNG_NET_STREAM_END': false,
+        'COUNTER_NET_SERVER_CONNECTION_CLOSE': false,
+        'DTRACE_NET_SERVER_CONNECTION': false,
+        'LTTNG_NET_SERVER_CONNECTION': false,
+        'COUNTER_NET_SERVER_CONNECTION': false
+      },
+
+    },
+    'plugins': {
+      promise
+    },
+    'rules': {
+      'promise/always-return': 'error',
+      'promise/no-return-wrap': 'error',
+      'promise/param-names': 'error',
+      'promise/catch-or-return': 'error',
+      'promise/no-native': 'off',
+      'promise/no-nesting': 'warn',
+      'promise/no-promise-in-callback': 'warn',
+      'promise/no-callback-in-promise': 'warn',
+      'promise/no-return-in-finally': 'warn',
+
+      // Possible Errors
+      // http://eslint.org/docs/rules/#possible-errors
+      'comma-dangle': [2, 'only-multiline'],
+      'no-control-regex': 2,
+      'no-debugger': 2,
+      'no-dupe-args': 2,
+      'no-dupe-keys': 2,
+      'no-duplicate-case': 2,
+      'no-empty-character-class': 2,
+      'no-ex-assign': 2,
+      'no-extra-boolean-cast': 2,
+      'no-extra-parens': [2, 'functions'],
+      'no-extra-semi': 2,
+      'no-func-assign': 2,
+      'no-invalid-regexp': 2,
+      'no-irregular-whitespace': 2,
+      'no-negated-in-lhs': 2,
+      'no-obj-calls': 2,
+      'no-proto': 2,
+      'no-unexpected-multiline': 2,
+      'no-unreachable': 2,
+      'use-isnan': 2,
+      'valid-typeof': 2,
+
+      // Best Practices
+      // http://eslint.org/docs/rules/#best-practices
+      'no-fallthrough': 2,
+      'no-octal': 2,
+      'no-redeclare': 2,
+      'no-self-assign': 2,
+      'no-unused-labels': 2,
+
+      // Strict Mode
+      // http://eslint.org/docs/rules/#strict-mode
+      'strict': [2, 'never'],
+
+      // Variables
+      // http://eslint.org/docs/rules/#variables
+      'no-delete-var': 2,
+      'no-undef': 2,
+      'no-unused-vars': [2, { 'args': 'none' }],
+
+      // Node.js and CommonJS
+      // http://eslint.org/docs/rules/#nodejs-and-commonjs
+      'no-mixed-requires': 2,
+      'no-new-require': 2,
+      'no-path-concat': 2,
+      'no-restricted-modules': [2, 'sys', '_linklist'],
+
+      // Stylistic Issues
+      // http://eslint.org/docs/rules/#stylistic-issues
+      'comma-spacing': 2,
+      'eol-last': 2,
+      'indent': [2, 2, { 'SwitchCase': 1 }],
+      'keyword-spacing': 2,
+      'max-len': [2, 120, 2],
+      'new-parens': 2,
+      'no-mixed-spaces-and-tabs': 2,
+      'no-multiple-empty-lines': [2, { 'max': 2 }],
+      'no-trailing-spaces': [2, { 'skipBlankLines': false }],
+      'quotes': [2, 'single', 'avoid-escape'],
+      'semi': 2,
+      'space-before-blocks': [2, 'always'],
+      'space-before-function-paren': [2, 'never'],
+      'space-in-parens': [2, 'never'],
+      'space-infix-ops': 2,
+      'space-unary-ops': 2,
+
+      // ECMAScript 6
+      // http://eslint.org/docs/rules/#ecmascript-6
+      'arrow-parens': [2, 'always'],
+      'arrow-spacing': [2, { 'before': true, 'after': true }],
+      'constructor-super': 2,
+      'no-class-assign': 2,
+      'no-confusing-arrow': 2,
+      'no-const-assign': 2,
+      'no-dupe-class-members': 2,
+      'no-new-symbol': 2,
+      'no-this-before-super': 2,
+      'prefer-const': 2
+    },
+    'ignores': []
+  }
+];

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}) {
   let key;
   logger = logger || noopLogger;
 
@@ -15,10 +36,18 @@ async function addFileToCache(client, logger, path,
       language: language || '',
       voice: voice || deploymentId,
       engine,
+      model,
       text,
     });
+    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

@@ -1,9 +1,12 @@
 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_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 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';
 
@@ -12,9 +15,11 @@ const HTTP_TIMEOUT = 5000;
 module.exports = {
   JAMBONES_TTS_TRIM_SILENCE,
   JAMBONES_DISABLE_TTS_STREAMING,
+  JAMBONES_DISABLE_AZURE_TTS_STREAMING,
   JAMBONES_HTTP_PROXY_IP,
   JAMBONES_HTTP_PROXY_PORT,
   JAMBONES_TTS_CACHE_DURATION_MINS,
+  JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
   TMP_FOLDER,
   HTTP_TIMEOUT
 };

lib/get-aws-sts-token.js

@@ -1,35 +1,58 @@
 const { STSClient, GetSessionTokenCommand, AssumeRoleCommand } = require('@aws-sdk/client-sts');
 const {makeAwsKey, noopLogger} = require('./utils');
 const debug = require('debug')('jambonz:speech-utils');
-const EXPIRY = 3600;
+const EXPIRY = process.env.AWS_STS_SESSION_DURATION || 3600;
+// by default reset aws session before expiry time 10 mins
+const CACHE_EXPIRY = process.env.AWS_STS_SESSION_RESET_EXPIRY || (EXPIRY - 600);
 
 async function getAwsAuthToken(
   logger, createHash, retrieveHash,
-  awsAccessKeyId, awsSecretAccessKey, awsRegion, roleArn = null) {
+  {speech_credential_sid, accessKeyId, secretAccessKey, region, roleArn}) {
   logger = logger || noopLogger;
   try {
-    const key = makeAwsKey(roleArn || awsAccessKeyId);
+    // 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: awsRegion});
+      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 {
+    } else if (accessKeyId) {
-      /* access token not found in cache, so generate it using STS */
       const stsClient = new STSClient({
-        region: awsRegion,
+        region,
         credentials: {
-          accessKeyId: awsAccessKeyId,
+          accessKeyId,
-          secretAccessKey: awsSecretAccessKey,
+          secretAccessKey,
         }
       });
       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 = {
@@ -38,10 +61,11 @@ async function getAwsAuthToken(
       sessionToken: data.Credentials.SessionToken,
       securityToken: data.Credentials.SessionToken
     };
-
+    // Only cache if expiry is good
-    /* expire 10 minutes before the hour, so we don't lose the use of it during a call */
+    if (expiry > 0) {
-    createHash(key, credentials, EXPIRY - 600)
+      createHash(key, credentials, expiry)
-      .catch((err) => logger.error(err, `Error saving hash for key ${key}`));
+        .catch((err) => logger.error(err, `Error saving hash for key ${key}`));
+    }
 
     return {...credentials, servedFromCache: false};
   } catch (err) {

lib/get-tts-voices.js

@@ -107,7 +107,12 @@ const getAwsVoices = async(_client, createHash, retrieveHash, logger, credential
     } else if (roleArn) {
       client = new PollyClient({
         region,
-        credentials: await getAwsAuthToken(logger, createHash, retrieveHash, null, null, region, roleArn),
+        credentials: await getAwsAuthToken(
+          logger, createHash, retrieveHash,
+          {
+            region,
+            roleArn
+          }),
       });
     } else {
       client = new PollyClient({region});

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} = {all: true}) {
   logger = logger || noopLogger;
 
   let purgedCount = 0, error;
@@ -33,6 +33,7 @@ async function purgeTtsCache(client, logger, {all, account_sid, vendor,
         language: language || '',
         voice: voice || deploymentId,
         engine,
+        model,
         text,
       });
       purgedCount = await client.del(key);

lib/synth-audio.js

@@ -3,6 +3,7 @@ const fs = require('fs');
 const bent = require('bent');
 const ttsGoogle = require('@google-cloud/text-to-speech');
 const { PollyClient, SynthesizeSpeechCommand } = require('@aws-sdk/client-polly');
+const { CartesiaClient } = require('@cartesia/cartesia-js');
 
 const sdk = require('microsoft-cognitiveservices-speech-sdk');
 const TextToSpeechV1 = require('ibm-watson/text-to-speech/v1');
@@ -20,7 +21,8 @@ const {
   createKryptonClient,
   createRivaClient,
   noopLogger,
-  makeFilePath
+  makeFilePath,
+  makePlayhtKey
 } = require('./utils');
 const getNuanceAccessToken = require('./get-nuance-access-token');
 const getVerbioAccessToken = require('./get-verbio-token');
@@ -40,9 +42,11 @@ const {AudioEncoding} = require('../stubs/riva/proto/riva_audio_pb');
 const debug = require('debug')('jambonz:realtimedb-helpers');
 const {
   JAMBONES_DISABLE_TTS_STREAMING,
+  JAMBONES_DISABLE_AZURE_TTS_STREAMING,
   JAMBONES_HTTP_PROXY_IP,
   JAMBONES_HTTP_PROXY_PORT,
   JAMBONES_TTS_CACHE_DURATION_MINS,
+  JAMBONES_TTS_TRIM_SILENCE,
 } = require('./config');
 const EXPIRES = JAMBONES_TTS_CACHE_DURATION_MINS;
 const OpenAI = require('openai');
@@ -85,15 +89,15 @@ const trimTrailingSilence = (buffer) => {
  */
 async function synthAudio(client, createHash, retrieveHash, logger, stats, { account_sid,
   vendor, language, voice, gender, text, engine, salt, model, credentials, deploymentId,
-  disableTtsCache, renderForCaching, disableTtsStreaming, options
+  disableTtsCache, renderForCaching = false, disableTtsStreaming, options
 }) {
-  let audioBuffer;
+  let audioData;
   let servedFromCache = false;
   let rtt;
   logger = logger || noopLogger;
 
   assert.ok(['google', 'aws', 'polly', 'microsoft', 'wellsaid', 'nuance', 'nvidia', 'ibm', 'elevenlabs',
-    'whisper', 'deepgram', 'playht', 'rimelabs', 'verbio'].includes(vendor) ||
+    'whisper', 'deepgram', 'playht', 'rimelabs', 'verbio', 'cartesia'].includes(vendor) ||
   vendor.startsWith('custom'),
   `synthAudio supported vendors are google, aws, microsoft, nuance, nvidia and wellsaid ..etc, not ${vendor}`);
   if ('google' === vendor) {
@@ -150,17 +154,26 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
     assert.ok(voice, 'synthAudio requires voice when verbio is used');
     assert.ok(credentials.client_id, 'synthAudio requires client_id when verbio is used');
     assert.ok(credentials.client_secret, 'synthAudio requires client_secret when verbio is used');
+  } else if ('deepgram' === vendor) {
+    if (!credentials.deepgram_tts_uri) {
+      assert.ok(credentials.api_key, 'synthAudio requires api_key when deepgram is used');
+    }
+  } else if ('cartesia' === vendor) {
+    assert.ok(credentials.api_key, 'synthAudio requires api_key when cartesia is used');
+    assert.ok(credentials.model_id, 'synthAudio requires model_id when cartesia is used');
   }
+
   const key = makeSynthKey({
     account_sid,
     vendor,
     language: language || '',
     voice: voice || deploymentId,
     engine,
+    // model or model_id is used to identify the tts cache.
+    model: model || credentials.model_id,
     text
   });
-  let filePath;
+
-  filePath = makeFilePath(vendor, key, salt);
   debug(`synth key is ${key}`);
   let cached;
   if (!disableTtsCache) {
@@ -171,10 +184,11 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
     debug('result WAS found in cache');
     servedFromCache = true;
     stats.increment('tts.cache.requests', ['found:yes']);
-    audioBuffer = Buffer.from(cached, 'base64');
+    audioData = JSON.parse(cached);
+    // convert base64 audio to buffer
+    audioData.audioContent = Buffer.from(audioData.audioContent, 'base64');
     client.expire(key, EXPIRES).catch((err) => logger.info(err, 'Error setting expires'));
-  }
+  } else {
-  if (!cached) {
     // not found in cache - go get it from speech vendor and add to cache
     debug('result was NOT found in cache');
     stats.increment('tts.cache.requests', ['found:no']);
@@ -182,82 +196,92 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
     const startAt = process.hrtime();
     switch (vendor) {
       case 'google':
-        audioBuffer = await synthGoogle(logger, {credentials, stats, language, voice, gender, text});
+        audioData = await synthGoogle(logger, {credentials, stats, language, voice, gender, text});
         break;
       case 'aws':
       case 'polly':
         vendorLabel = 'aws';
-        audioBuffer = await synthPolly(createHash, retrieveHash, logger,
+        audioData = await synthPolly(createHash, retrieveHash, logger,
           {credentials, stats, language, voice, text, engine});
         break;
       case 'azure':
       case 'microsoft':
         vendorLabel = 'microsoft';
-        audioBuffer = await synthMicrosoft(logger, {credentials, stats, language, voice, text, deploymentId,
+        audioData = await synthMicrosoft(logger, {credentials, stats, language, voice, text, deploymentId,
-          filePath, renderForCaching, disableTtsStreaming});        
+          renderForCaching, disableTtsStreaming});
         break;
       case 'nuance':
         model = model || 'enhanced';
-        audioBuffer = await synthNuance(client, logger, {credentials, stats, voice, model, text});
+        audioData = await synthNuance(client, logger, {credentials, stats, voice, model, text});
         break;
       case 'nvidia':
-        audioBuffer = await synthNvidia(client, logger, {credentials, stats, language, voice, model, text});
+        audioData = await synthNvidia(client, logger, {credentials, stats, language, voice, model, text});
         break;
       case 'ibm':
-        audioBuffer = await synthIbm(logger, {credentials, stats, voice, text});
+        audioData = await synthIbm(logger, {credentials, stats, voice, text});
         break;
       case 'wellsaid':
-        audioBuffer = await synthWellSaid(logger, {credentials, stats, language, voice, text, filePath});
+        audioData = await synthWellSaid(logger, {credentials, stats, language, voice, text});
         break;
       case 'elevenlabs':
-        audioBuffer = await synthElevenlabs(logger, {
+        audioData = await synthElevenlabs(logger, {
-          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming, filePath
+          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming});
-        });        
         break;
       case 'playht':
-        audioBuffer = await synthPlayHT(logger, {
+        audioData = await synthPlayHT(client, logger, {
-          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming, filePath
+          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming});
-        });        
+        break;
+      case 'cartesia':
+        audioData = await synthCartesia(logger, {
+          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming});
         break;
       case 'rimelabs':
-        audioBuffer = await synthRimelabs(logger, {
+        audioData = await synthRimelabs(logger, {
-          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming, filePath
+          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming});
-        });        
         break;
       case 'whisper':
-        audioBuffer = await synthWhisper(logger, {
+        audioData = await synthWhisper(logger, {
-          credentials, stats, voice, text, renderForCaching, disableTtsStreaming});        
+          credentials, stats, voice, text, renderForCaching, disableTtsStreaming});
         break;
       case 'verbio':
-        audioBuffer = await synthVerbio(client, logger, {
+        audioData = await synthVerbio(client, logger, {
           credentials, stats, voice, text, renderForCaching, disableTtsStreaming});
-        if (audioBuffer?.filePath) return audioBuffer;
+        if (audioData?.filePath) return audioData;
         break;
       case 'deepgram':
-        audioBuffer = await synthDeepgram(logger, {credentials, stats, model, text,
+        audioData = await synthDeepgram(logger, {credentials, stats, model, text,
-          renderForCaching, disableTtsStreaming});        
+          renderForCaching, disableTtsStreaming});
         break;
       case vendor.startsWith('custom') ? vendor : 'cant_match_value':
-        ({ audioBuffer, filePath } = await synthCustomVendor(logger,
+        audioData = await synthCustomVendor(logger,
-          {credentials, stats, language, voice, text, filePath}));
+          {credentials, stats, language, voice, text});
         break;
       default:
         assert(`synthAudio: unsupported speech vendor ${vendor}`);
     }
-    if('filePath' in audioBuffer) return audioBuffer;
+    if ('filePath' in audioData) return audioData;
     const diff = process.hrtime(startAt);
     const time = diff[0] * 1e3 + diff[1] * 1e-6;
     rtt = time.toFixed(0);
     stats.histogram('tts.response_time', rtt, [`vendor:${vendorLabel}`]);
     debug(`tts rtt time for ${text.length} chars on ${vendorLabel}: ${rtt}`);
     logger.info(`tts rtt time for ${text.length} chars on ${vendorLabel}: ${rtt}`);
-
+    // Save audio json to cache
-    client.setex(key, EXPIRES, audioBuffer.toString('base64'))
+    client.setex(key, EXPIRES, JSON.stringify({
+      ...audioData,
+      audioContent: audioData.audioContent?.toString('base64')
+    }))
       .catch((err) => logger.error(err, `error calling setex on key ${key}`));
   }
 
   return new Promise((resolve, reject) => {
-    fs.writeFile(filePath, audioBuffer, (err) => {
+    const { audioContent, extension } = audioData;
+    const filePath = makeFilePath({
+      key,
+      salt,
+      extension
+    });
+    fs.writeFile(filePath, audioContent, (err) => {
       if (err) return reject(err);
       resolve({filePath, servedFromCache, rtt});
     });
@@ -280,7 +304,12 @@ const synthPolly = async(createHash, retrieveHash, logger,
     } else if (roleArn) {
       polly = new PollyClient({
         region,
-        credentials: await getAwsAuthToken(logger, createHash, retrieveHash, null, null, region, roleArn),
+        credentials: await getAwsAuthToken(
+          logger, createHash, retrieveHash,
+          {
+            region,
+            roleArn
+          }),
       });
     } else {
       // AWS RoleArn assigned to Instance profile
@@ -307,7 +336,13 @@ const synthPolly = async(createHash, retrieveHash, logger,
         .on('data', (chunk) => {
           chunks.push(chunk);
         })
-        .on('end', () => resolve(Buffer.concat(chunks)));
+        .on('end', () => resolve(
+          {
+            audioContent: Buffer.concat(chunks),
+            extension: 'mp3',
+            sampleRate: 8000
+          }
+        ));
     });
   } catch (err) {
     logger.info({err}, 'synthAudio: Error synthesizing speech using aws polly');
@@ -318,6 +353,48 @@ 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 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;
+    }
+  }
+
   const opts = {
     voice: {
       ...(typeof voice === 'string' && {name: voice}),
@@ -332,7 +409,11 @@ const synthGoogle = async(logger, {credentials, stats, language, voice, gender,
     const responses = await client.synthesizeSpeech(opts);
     stats.increment('tts.count', ['vendor:google', 'accepted:yes']);
     client.close();
-    return responses[0].audioContent;
+    return {
+      audioContent: responses[0].audioContent,
+      extension: 'mp3',
+      sampleRate: 8000
+    };
   } catch (err) {
     console.error(err);
     logger.info({err, opts}, 'synthAudio: Error synthesizing speech using google');
@@ -363,7 +444,11 @@ const synthIbm = async(logger, {credentials, stats, voice, text}) => {
     for await (const chunk of r.result) {
       chunks.push(chunk);
     }
-    return Buffer.concat(chunks);
+    return {
+      audioContent: Buffer.concat(chunks),
+      extension: 'mp3',
+      sampleRate: 8000
+    };
   } catch (err) {
     logger.info({err, params}, 'synthAudio: Error synthesizing speech using ibm');
     stats.increment('tts.count', ['vendor:ibm', 'accepted:no']);
@@ -373,11 +458,9 @@ const synthIbm = async(logger, {credentials, stats, voice, text}) => {
 
 async function _synthOnPremMicrosoft(logger, {
   credentials,
-  stats,
   language,
   voice,
-  text,
+  text
-  filePath
 }) {
   const {use_custom_tts, custom_tts_endpoint_url, api_key} = credentials;
   let content = text;
@@ -399,15 +482,19 @@ async function _synthOnPremMicrosoft(logger, {
   }
 
   try {
-    const trimSilence = filePath.endsWith('.r8');
+    const trimSilence = JAMBONES_TTS_TRIM_SILENCE;
     const post = bent('POST', 'buffer', {
       'X-Microsoft-OutputFormat': trimSilence ? 'raw-8khz-16bit-mono-pcm' : 'audio-16khz-32kbitrate-mono-mp3',
       'Content-Type': 'application/ssml+xml',
       'User-Agent': 'Jambonz',
       ...(api_key && {'Ocp-Apim-Subscription-Key': api_key})
     });
-    const mp3 = await post(custom_tts_endpoint_url, content);
+    const audioContent = await post(custom_tts_endpoint_url, content);
-    return mp3;
+    return {
+      audioContent,
+      extension: trimSilence ? 'r8' : 'mp3',
+      sampleRate: 8000
+    };
   } catch (err) {
     logger.info({err}, '_synthMicrosoftByHttp returned error');
     throw err;
@@ -420,7 +507,6 @@ const synthMicrosoft = async(logger, {
   language,
   voice,
   text,
-  filePath,
   renderForCaching,
   disableTtsStreaming
 }) => {
@@ -443,7 +529,8 @@ const synthMicrosoft = async(logger, {
       content = `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="${language}"><voice name="${voice}">${words}</voice></speak>`;
       logger.info({content}, 'synthMicrosoft');
     }
-    if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
+    if (!JAMBONES_DISABLE_TTS_STREAMING && !JAMBONES_DISABLE_AZURE_TTS_STREAMING &&
+      !renderForCaching && !disableTtsStreaming) {
       let params = '';
       params += `{api_key=${apiKey}`;
       params += `,language=${language}`;
@@ -462,17 +549,18 @@ const synthMicrosoft = async(logger, {
         rtt: 0
       };
     }
+    // Azure Onprem
     if (use_custom_tts && custom_tts_endpoint_url) {
       return await _synthOnPremMicrosoft(logger, {
         credentials,
         stats,
         language,
         voice,
-        text,
+        text
-        filePath
       });
     }
-    const trimSilence = filePath.endsWith('.r8');
+    // Azure hosted service
+    const trimSilence = JAMBONES_TTS_TRIM_SILENCE;
     const speechConfig = SpeechConfig.fromSubscription(apiKey, region);
     speechConfig.speechSynthesisLanguage = language;
     speechConfig.speechSynthesisVoiceName = voice;
@@ -507,7 +595,11 @@ const synthMicrosoft = async(logger, {
             case ResultReason.SynthesizingAudioCompleted:
               let buffer = Buffer.from(result.audioData);
               if (trimSilence) buffer = trimTrailingSilence(buffer);
-              resolve(buffer);
+              resolve({
+                audioContent: buffer,
+                extension: trimSilence ? 'r8' : 'mp3',
+                sampleRate: 8000
+              });
               synthesizer.close();
               stats.increment('tts.count', ['vendor:microsoft', 'accepted:yes']);
               break;
@@ -537,11 +629,15 @@ const synthWellSaid = async(logger, {credentials, stats, language, voice, gender
       'Accept': 'audio/mpeg',
       'Content-Type': 'application/json'
     });
-    const mp3 = await post('/v1/tts/stream', {
+    const audioContent = await post('/v1/tts/stream', {
       text,
       speaker_id: voice
     });
-    return mp3;
+    return {
+      audioContent,
+      extension: 'mp3',
+      sampleRate: 8000
+    };
   } catch (err) {
     logger.info({err}, 'testWellSaidTts returned error');
     throw err;
@@ -578,8 +674,8 @@ const synthNuance = async(client, logger, {credentials, stats, voice, model, tex
     t.setText(text);
     input.setText(t);
   }
-
+  const sampleRate = 8000;
-  pcm.setSampleRateHz(8000);
+  pcm.setSampleRateHz(sampleRate);
   f.setPcm(pcm);
   p.setAudioFormat(f);
   v.setName(voice);
@@ -603,7 +699,11 @@ const synthNuance = async(client, logger, {credentials, stats, voice, model, tex
         const details = status.getDetails();
         return reject({code, message, details});
       }
-      resolve(Buffer.from(response.getAudio()));
+      resolve({
+        audioContent: Buffer.from(response.getAudio()),
+        extension: 'r8',
+        sampleRate
+      });
     });
   });
 };
@@ -611,12 +711,13 @@ const synthNuance = async(client, logger, {credentials, stats, voice, model, tex
 const synthNvidia = async(client, logger, {credentials, stats, language,  voice, model, text}) => {
   const {riva_server_uri} = credentials;
   let rivaClient, request;
+  const sampleRate = 8000;
   try {
     rivaClient = await createRivaClient(riva_server_uri);
     request = new SynthesizeSpeechRequest();
     request.setVoiceName(voice);
     request.setLanguageCode(language);
-    request.setSampleRateHz(8000);
+    request.setSampleRateHz(sampleRate);
     request.setEncoding(AudioEncoding.LINEAR_PCM);
     request.setText(text);
   } catch (err) {
@@ -630,7 +731,11 @@ const synthNvidia = async(client, logger, {credentials, stats, language,  voice,
         logger.info({err, voice, language}, 'error synthesizing speech using Nvidia');
         return reject(err);
       }
-      resolve(Buffer.from(response.getAudio()));
+      resolve({
+        audioContent: Buffer.from(response.getAudio()),
+        extension: 'r8',
+        sampleRate
+      });
     });
   });
 };
@@ -652,12 +757,13 @@ const synthCustomVendor = async(logger, {credentials, stats, language, voice, te
       text
     });
 
-    const regex = /\.[^\.]*$/g;
     const mime = response.headers['content-type'];
     const buffer = await response.arrayBuffer();
+    const [extension, sampleRate] = getFileExtFromMime(mime);
     return {
-      audioBuffer: buffer,
+      audioContent: buffer,
-      filePath: filePath.replace(regex, getFileExtFromMime(mime))
+      extension,
+      sampleRate
     };
   } catch (err) {
     logger.info({err}, `Vendor ${vendor} returned error`);
@@ -684,6 +790,8 @@ const synthElevenlabs = async(logger, {
     if (opts.voice_settings?.stability) params += `,stability=${opts.voice_settings.stability}`;
     if (opts.voice_settings?.style) params += `,style=${opts.voice_settings.style}`;
     if (opts.voice_settings?.use_speaker_boost === false) params += ',use_speaker_boost=false';
+    if (opts.previous_text) params += `,previous_text=${opts.previous_text}`;
+    if (opts.next_text) params += `,next_text=${opts.next_text}`;
     params += '}';
 
     return {
@@ -701,7 +809,7 @@ const synthElevenlabs = async(logger, {
       'Accept': 'audio/mpeg',
       'Content-Type': 'application/json'
     });
-    const mp3 = await post(`/v1/text-to-speech/${voice}${optimize_streaming_latency}`, {
+    const audioContent = await post(`/v1/text-to-speech/${voice}${optimize_streaming_latency}`, {
       text,
       model_id,
       voice_settings: {
@@ -710,7 +818,11 @@ const synthElevenlabs = async(logger, {
       },
       ...opts
     });
-    return mp3;
+    return {
+      audioContent,
+      extension: 'mp3',
+      sampleRate: 8000
+    };
   } catch (err) {
     logger.info({err}, 'synth Elevenlabs returned error');
     stats.increment('tts.count', ['vendor:elevenlabs', 'accepted:no']);
@@ -718,12 +830,40 @@ const synthElevenlabs = async(logger, {
   }
 };
 
-const synthPlayHT = async(logger, {
+const synthPlayHT = async(client, logger, {
-  credentials, options, stats, voice, text, renderForCaching, disableTtsStreaming
+  credentials, options, stats, voice, language, text, renderForCaching, disableTtsStreaming
 }) => {
   const {api_key, user_id, voice_engine, options: credOpts} = credentials;
   const opts = !!options && Object.keys(options).length !== 0 ? options : JSON.parse(credOpts || '{}');
 
+  let synthesizeUrl = 'https://api.play.ht/api/v2/tts/stream';
+
+  // If model is play3.0, the synthesizeUrl is got from authentication endpoint
+  if (voice_engine === 'Play3.0') {
+    try {
+      const post = bent('https://api.play.ht', 'POST', 'json', 201, {
+        'AUTHORIZATION': api_key,
+        'X-USER-ID': user_id,
+        'Accept': 'application/json'
+      });
+      const key = makePlayhtKey(api_key);
+      const url = await client.get(key);
+      if (!url) {
+        const {inference_address, expires_at_ms} = await post('/api/v3/auth');
+        synthesizeUrl = inference_address;
+        const expiry =  Math.floor((expires_at_ms - Date.now()) / 1000 - 30);
+        await client.set(key, inference_address, 'EX', expiry);
+      } else {
+        // Use cached URL
+        synthesizeUrl = url;
+      }
+    } catch (err) {
+      logger.info({err}, 'synth PlayHT returned error for authentication version 3.0');
+      stats.increment('tts.count', ['vendor:playht', 'accepted:no']);
+      throw err;
+    }
+  }
+
   /* default to using the streaming interface, unless disabled by env var OR we want just a cache file */
   if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
     let params = '';
@@ -732,7 +872,9 @@ const synthPlayHT = async(logger, {
     params += ',vendor=playht';
     params += `,voice=${voice}`;
     params += `,voice_engine=${voice_engine}`;
+    params += `,synthesize_url=${synthesizeUrl}`;
     params += ',write_cache_file=1';
+    params += `,language=${language}`;
     if (opts.quality) params += `,quality=${opts.quality}`;
     if (opts.speed) params += `,speed=${opts.speed}`;
     if (opts.seed) params += `,style=${opts.seed}`;
@@ -741,6 +883,8 @@ const synthPlayHT = async(logger, {
     if (opts.voice_guidance) params += `,voice_guidance=${opts.voice_guidance}`;
     if (opts.style_guidance) params += `,style_guidance=${opts.style_guidance}`;
     if (opts.text_guidance) params += `,text_guidance=${opts.text_guidance}`;
+    if (opts.top_p) params += `,top_p=${opts.top_p}`;
+    if (opts.repetition_penalty) params += `,repetition_penalty=${opts.repetition_penalty}`;
     params += '}';
 
     return {
@@ -751,21 +895,29 @@ const synthPlayHT = async(logger, {
   }
 
   try {
-    const post = bent('https://api.play.ht', 'POST', 'buffer', {
+    const post = bent('POST', 'buffer', {
-      'AUTHORIZATION': api_key,
+      ...(voice_engine !== 'Play3.0' && {
-      'X-USER-ID': user_id,
+        'AUTHORIZATION': api_key,
+        'X-USER-ID': user_id,
+      }),
       'Accept': 'audio/mpeg',
       'Content-Type': 'application/json'
     });
-    const mp3 = await post('/api/v2/tts/stream', {
+
+    const audioContent = await post(synthesizeUrl, {
       text,
+      ...(voice_engine === 'Play3.0' && { language }),
       voice,
       voice_engine,
       output_format: 'mp3',
       sample_rate: 8000,
       ...opts
     });
-    return mp3;
+    return {
+      audioContent,
+      extension: 'mp3',
+      sampleRate: 8000
+    };
   } catch (err) {
     logger.info({err}, 'synth PlayHT returned error');
     stats.increment('tts.count', ['vendor:playht', 'accepted:no']);
@@ -774,7 +926,7 @@ const synthPlayHT = async(logger, {
 };
 
 const synthRimelabs = async(logger, {
-  credentials, options, stats, voice, text, renderForCaching, disableTtsStreaming
+  credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming
 }) => {
   const {api_key, model_id, options: credOpts} = credentials;
   const opts = !!options && Object.keys(options).length !== 0 ? options : JSON.parse(credOpts || '{}');
@@ -785,6 +937,7 @@ const synthRimelabs = async(logger, {
     params += `{api_key=${api_key}`;
     params += `,model_id=${model_id}`;
     params += ',vendor=rimelabs';
+    params += `,language=${language}`;
     params += `,voice=${voice}`;
     params += ',write_cache_file=1';
     if (opts.speedAlpha) params += `,speed_alpha=${opts.speedAlpha}`;
@@ -804,14 +957,20 @@ const synthRimelabs = async(logger, {
       'Accept': 'audio/mp3',
       'Content-Type': 'application/json'
     });
-    const mp3 = await post('/v1/rime-tts', {
+    const sampleRate = 8000;
+    const audioContent = await post('/v1/rime-tts', {
       speaker: voice,
       text,
       modelId: model_id,
-      samplingRate: 8000,
+      samplingRate: sampleRate,
+      lang: language,
       ...opts
     });
-    return mp3;
+    return {
+      audioContent,
+      extension: 'mp3',
+      sampleRate
+    };
   } catch (err) {
     logger.info({err}, 'synth rimelabs returned error');
     stats.increment('tts.count', ['vendor:rimelabs', 'accepted:no']);
@@ -845,13 +1004,17 @@ const synthVerbio = async(client, logger, {credentials, stats, voice, text, rend
       'User-Agent': 'jambonz',
       'Content-Type': 'application/json'
     });
-    const r8 = await post('/api/v1/synthesize', {
+    const audioContent = await post('/api/v1/synthesize', {
       voice_id: voice,
       output_sample_rate: '8k',
       output_encoding: 'pcm16',
       text
     });
-    return r8;
+    return {
+      audioContent,
+      extension: 'r8',
+      sampleRate: 8000
+    };
   } catch (err) {
     logger.info({err}, 'synth Verbio returned error');
     stats.increment('tts.count', ['vendor:verbio', 'accepted:no']);
@@ -891,7 +1054,11 @@ const synthWhisper = async(logger, {credentials, stats, voice, text, renderForCa
       input: text,
       response_format: 'mp3'
     });
-    return Buffer.from(await mp3.arrayBuffer());
+    return {
+      audioContent: Buffer.from(await mp3.arrayBuffer()),
+      extension: 'mp3',
+      sampleRate: 8000
+    };
   } catch (err) {
     logger.info({err}, 'synth whisper returned error');
     stats.increment('tts.count', ['vendor:openai', 'accepted:no']);
@@ -900,13 +1067,14 @@ const synthWhisper = async(logger, {credentials, stats, voice, text, renderForCa
 };
 
 const synthDeepgram = async(logger, {credentials, stats, model, text, renderForCaching, disableTtsStreaming}) => {
-  const {api_key} = credentials;
+  const {api_key, deepgram_tts_uri} = credentials;
   if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
     let params = '';
     params += `{api_key=${api_key}`;
     params += ',vendor=deepgram';
     params += `,voice=${model}`;
     params += ',write_cache_file=1';
+    if (deepgram_tts_uri) params += `,endpoint=${deepgram_tts_uri}`;
     params += '}';
 
     return {
@@ -916,15 +1084,20 @@ const synthDeepgram = async(logger, {credentials, stats, model, text, renderForC
     };
   }
   try {
-    const post = bent('https://api.deepgram.com', 'POST', 'buffer', {
+    const post = bent(deepgram_tts_uri || 'https://api.deepgram.com', 'POST', 'buffer', {
-      'Authorization': `Token ${api_key}`,
+      // on-premise deepgram does not  require to have api_key
+      ...(api_key && {'Authorization': `Token ${api_key}`}),
       'Accept': 'audio/mpeg',
       'Content-Type': 'application/json'
     });
-    const mp3 = await post(`/v1/speak?model=${model}`, {
+    const audioContent = await post(`/v1/speak?model=${model}`, {
       text
     });
-    return mp3;
+    return {
+      audioContent,
+      extension: 'mp3',
+      sampleRate: 8000
+    };
   } catch (err) {
     logger.info({err}, 'synth Deepgram returned error');
     stats.increment('tts.count', ['vendor:deepgram', 'accepted:no']);
@@ -932,26 +1105,96 @@ const synthDeepgram = async(logger, {credentials, stats, model, text, renderForC
   }
 };
 
+const synthCartesia = async(logger, {
+  credentials, options, stats, voice, language, text, renderForCaching, disableTtsStreaming
+}) => {
+  const {api_key, model_id, embedding, options: credOpts} = credentials;
+  const opts = !!options && Object.keys(options).length !== 0 ? options : JSON.parse(credOpts || '{}');
+
+  if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
+    let params = '';
+    params += `{api_key=${api_key}`;
+    params += `,model_id=${model_id}`;
+    params += ',vendor=cartesia';
+    params += `,voice=${voice}`;
+    params += ',write_cache_file=1';
+    params += `,language=${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}`;
+    params += '}';
+
+    return {
+      filePath: `say:${params}${text.replace(/\n/g, ' ').replace(/\r/g, ' ')}`,
+      servedFromCache: false,
+      rtt: 0
+    };
+  }
+
+  try {
+    const client = new CartesiaClient({ apiKey: api_key });
+    const sampleRate = 48000;
+    const mp3 = await client.tts.bytes({
+      modelId: model_id,
+      transcript: text,
+      voice: {
+        mode: embedding ? 'embedding' : 'id',
+        ...(embedding ?
+          {
+            embedding: embedding.split(',').map(Number)
+          } :
+          {
+            id: voice
+          }
+        ),
+        ...(opts.speed || opts.emotion && {
+          experimentalControls: {
+            ...(opts.speed !== null && opts.speed !== undefined && {speed: opts.speed}),
+            ...(opts.emotion && {emotion: opts.emotion}),
+          }
+        })
+      },
+      language: language,
+      outputFormat: {
+        container: 'mp3',
+        bitRate: 128000,
+        sampleRate
+      },
+    });
+    return {
+      audioContent: Buffer.from(mp3),
+      extension: 'mp3',
+      sampleRate
+    };
+  } catch (err) {
+    logger.info({err}, 'synth Cartesia returned error');
+    stats.increment('tts.count', ['vendor:cartesia', 'accepted:no']);
+    throw err;
+  }
+
+};
+
 const getFileExtFromMime = (mime) => {
   switch (mime) {
     case 'audio/wav':
     case 'audio/x-wav':
-      return '.wav';
+      return ['wav', 8000];
     case /audio\/l16.*rate=8000/.test(mime) ? mime : 'cant match value':
-      return '.r8';
+      return ['r8', 8000];
     case /audio\/l16.*rate=16000/.test(mime) ? mime : 'cant match value':
-      return '.r16';
+      return ['r16', 16000];
     case /audio\/l16.*rate=24000/.test(mime) ? mime : 'cant match value':
-      return '.r24';
+      return ['r24', 24000];
     case /audio\/l16.*rate=32000/.test(mime) ? mime : 'cant match value':
-      return '.r32';
+      return ['r32', 32000];
     case /audio\/l16.*rate=48000/.test(mime) ? mime : 'cant match value':
-      return '.r48';
+      return ['r48', 48000];
     case 'audio/mpeg':
     case 'audio/mp3':
-      return '.mp3';
+      return ['mp3', 8000];
     default:
-      return '.wav';
+      return ['wav', 8000];
   }
 };
 

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');
 /**
@@ -16,48 +16,27 @@ const debug = require('debug')('jambonz:realtimedb-helpers');
  */
 //const nuanceClientMap = new Map();
 
-function makeSynthKey({account_sid = '', vendor, language, voice, engine = '', text}) {
+function makeSynthKey({
+  account_sid = '',
+  vendor,
+  language,
+  voice,
+  engine = '',
+  model = '',
+  text
+}) {
   const hash = crypto.createHash('sha1');
-  hash.update(`${language}:${vendor}:${voice}:${engine}:${text}`);
+  hash.update(`${language}:${vendor}:${voice}:${engine}:${model}:${text}`);
   const hexHashKey = hash.digest('hex');
   const accountKey = account_sid ? `:${account_sid}` : '';
-  const namespace = vendor.startsWith('custom') ? vendor : getFileExtension(vendor);
+  const key = `tts${accountKey}:${hexHashKey}`;
-  const key = `tts${accountKey}:${namespace}:${hexHashKey}`;
   return key;
 }
 
-function makeFilePath(vendor, key, salt = '') {
+function makeFilePath({key, salt = '', extension}) {
-  const extension = getFileExtension(vendor);
   return `${TMP_FOLDER}/${key.replace('tts:', `tts-${salt}`)}.${extension}`;
 }
 
-function getFileExtension(vendor) {
-  const mp3Extension = 'mp3';
-  const r8Extension = 'r8';
-
-  switch (vendor) {
-    case 'azure':
-    case 'microsoft':
-      if (!JAMBONES_DISABLE_TTS_STREAMING || JAMBONES_TTS_TRIM_SILENCE) {
-        return r8Extension;
-      } else {
-        return mp3Extension;
-      }
-    case 'deepgram':
-    case 'elevenlabs':
-    case 'rimlabs':
-      if (!JAMBONES_DISABLE_TTS_STREAMING) {
-        return r8Extension;
-      } else {
-        return mp3Extension;
-      }
-    case 'nuance':
-    case 'nvidia':
-      return r8Extension;
-    default:
-      return mp3Extension;
-  }
-}
 
 const noopLogger = {
   info: () => {},
@@ -86,6 +65,11 @@ 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);
@@ -159,6 +143,7 @@ module.exports = {
   makeSynthKey,
   makeNuanceKey,
   makeIbmKey,
+  makePlayhtKey,
   makeAwsKey,
   makeVerbioKey,
   getNuanceAccessToken,

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/speech-utils",
-  "version": "0.1.2",
+  "version": "0.2.4",
   "description": "TTS-related speech utilities for jambonz",
   "main": "index.js",
   "author": "Dave Horton",
@@ -12,8 +12,9 @@
     "test": "NODE_ENV=test node test/ ",
     "coverage": "nyc --reporter html --report-dir ./coverage npm run test",
     "jslint": "eslint index.js lib",
-    "jslint:fix": "eslint --fix '**/*.js'",
+    "jslint:fix": "npm run jslint --fix",
-    "build": "./build_stubs.sh"
+    "build": "./build_stubs.sh",
+    "prepare": "husky"
   },
   "repository": {
     "type": "git",
@@ -27,7 +28,8 @@
   "dependencies": {
     "@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": "^2.1.0",
+    "@google-cloud/text-to-speech": "^5.5.0",
     "@grpc/grpc-js": "^1.9.14",
     "@jambonz/realtimedb-helpers": "^0.8.7",
     "bent": "^7.3.12",
@@ -35,16 +37,17 @@
     "form-urlencoded": "^6.1.4",
     "google-protobuf": "^3.21.2",
     "ibm-watson": "^8.0.0",
-    "microsoft-cognitiveservices-speech-sdk": "1.36.0",
+    "microsoft-cognitiveservices-speech-sdk": "1.38.0",
     "openai": "^4.25.0",
-    "undici": "^6.4.0"
+    "undici": "^7.5.0"
   },
   "devDependencies": {
-    "config": "^3.3.10",
+    "config": "^3.3.11",
-    "eslint": "^8.56.0",
+    "eslint": "^9.3.0",
-    "eslint-plugin-promise": "^6.1.1",
+    "eslint-plugin-promise": "^6.2.0",
+    "husky": "^9.0.11",
     "nyc": "^15.1.0",
-    "pino": "^8.17.0",
+    "pino": "^9.1.0",
-    "tape": "^5.7.3"
+    "tape": "^5.7.5"
   }
 }

test/aws.js

@@ -19,12 +19,20 @@ test('AWS - create and cache auth token', async(t) => {
       return;
   }
   try {
-    let obj = await getAwsAuthToken(process.env.AWS_ACCESS_KEY_ID, process.env.AWS_SECRET_ACCESS_KEY, process.env.AWS_REGION);
+    let obj = await getAwsAuthToken({
+      accessKeyId: process.env.AWS_ACCESS_KEY_ID, 
+      secretAccessKey: process.env.AWS_SECRET_ACCESS_KEY,
+      region: process.env.AWS_REGION
+    });
     //console.log({obj}, 'received auth token from AWS');
     t.ok(obj.securityToken && !obj.servedFromCache, 'successfullY generated auth token from AWS');
 
     await sleep(250);
-    obj = await getAwsAuthToken(process.env.AWS_ACCESS_KEY_ID, process.env.AWS_SECRET_ACCESS_KEY, process.env.AWS_REGION);
+    obj = await getAwsAuthToken({
+      accessKeyId: process.env.AWS_ACCESS_KEY_ID, 
+      secretAccessKey: process.env.AWS_SECRET_ACCESS_KEY,
+      region: process.env.AWS_REGION
+    });
     //console.log({obj}, 'received auth token from AWS - second request');
     t.ok(obj.securityToken && obj.servedFromCache, 'successfully received access token from cache');
  

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);
@@ -91,14 +91,17 @@ 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) {
+  if (!process.env.GCP_CUSTOM_VOICE_FILE &&
-    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');
+    !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: {
@@ -109,7 +112,7 @@ 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
       }
     });
@@ -121,6 +124,48 @@ 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
+      }
+    });
+    t.ok(!opts.servedFromCache, `successfully synthesized google voice cloning audio to ${opts.filePath}`);
+  } catch (err) {
+    console.error(err);
+    t.end(err);
+  }
+  client.quit();
+});
+
 test('AWS speech synth tests', async(t) => {
   const fn = require('..');
   const {synthAudio, client} = fn(opts, logger);
@@ -509,6 +554,7 @@ test('Custom Vendor speech synth tests', async(t) => {
       text: 'This is a test.  This is only a test',
     });
     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 +562,21 @@ 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',
+    });
+    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: {
@@ -574,9 +635,9 @@ 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);
 
@@ -584,32 +645,74 @@ 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
     });
     t.ok(!opts.servedFromCache, `successfully playht eleven audio to ${opts.filePath}`);
 
+  } catch (err) {
+    console.error(JSON.stringify(err));
+    t.end(err);
+  }
+  client.quit();
+};
+
+test('PlayHT speech synth tests', async(t) => {
+  await testPlayHT(t, 'PlayHT2.0-turbo');
+});
+
+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 playht eleven audio to ${opts.filePath}`);
+
   } catch (err) {
     console.error(JSON.stringify(err));
     t.end(err);
@@ -617,7 +720,7 @@ test('PlayHT speech synth tests', async(t) => {
   client.quit();
 });
 
-test('rimelabs speech synth tests', async(t) => {
+test('rimelabs speech synth tests mist', async(t) => {
   const fn = require('..');
   const {synthAudio, client} = fn(opts, logger);
 
@@ -627,7 +730,7 @@ 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 +740,7 @@ test('rimelabs speech synth tests', async(t) => {
           reduceLatency: false
         })
       },
-      language: 'en-US',
+      language: 'eng',
       voice: 'amber',
       text,
       renderForCaching: true
@@ -651,6 +754,40 @@ 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 +887,11 @@ 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}), 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`);
 
@@ -785,24 +922,15 @@ 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(purgedCountWhenErrored === 0, 'purged no records when specified key was not found');
-    t.ok(error, `error returned 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:*')
+    const cached = await client.keys('tts:*');
-    t.ok(cached.length >= 1, `successfully kept all non-specified tts records in cache`);
+    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    
+    process.env.VG_TRIM_TTS_SILENCE = 'true';
-    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';    
     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);