fix deepgram and playht has white noise from begining of cache audio

lib/add-file-to-cache.js

@@ -3,29 +3,8 @@ 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, model, text, instructions}) {
+  {account_sid, vendor, language, voice, deploymentId, engine, text}) {
   let key;
   logger = logger || noopLogger;
 
@@ -36,19 +15,10 @@ 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, JSON.stringify(
-      {
-        audioContent: audioBuffer.toString('base64'),
-        extension,
-        sampleRate
-      }
-    ));
+    await client.setex(key, EXPIRES, audioBuffer.toString('base64'));
   } catch (err) {
     logger.error(err, 'addFileToCache: Error');
     return;

lib/config.js

@@ -2,14 +2,13 @@ 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 = process.env.JAMBONES_TMP_FOLDER || '/tmp';
+const TMP_FOLDER = '/tmp';
 
 const HTTP_TIMEOUT = 5000;
 
@@ -22,6 +21,5 @@ module.exports = {
   JAMBONES_TTS_CACHE_DURATION_MINS,
   JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
   TMP_FOLDER,
-  HTTP_TIMEOUT,
-  JAMBONES_AZURE_ENABLE_SSML
+  HTTP_TIMEOUT
 };

lib/get-aws-sts-token.js

@@ -7,23 +7,22 @@ const CACHE_EXPIRY = process.env.AWS_STS_SESSION_RESET_EXPIRY || (EXPIRY - 600);
 
 async function getAwsAuthToken(
   logger, createHash, retrieveHash,
-  {speech_credential_sid, accessKeyId, secretAccessKey, region, roleArn}) {
+  {accessKeyId, secretAccessKey, region, roleArn}) {
   logger = logger || noopLogger;
   try {
-    // 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 key = makeAwsKey(roleArn || accessKeyId);
     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 if (accessKeyId) {
+    } else {
+      /* access token not found in cache, so generate it using STS */
       const stsClient = new STSClient({
         region,
         credentials: {
@@ -33,26 +32,6 @@ 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 = {
@@ -61,11 +40,9 @@ async function getAwsAuthToken(
       sessionToken: data.Credentials.SessionToken,
       securityToken: data.Credentials.SessionToken
     };
-    // Only cache if expiry is good
-    if (expiry > 0) {
-      createHash(key, credentials, expiry)
-        .catch((err) => logger.error(err, `Error saving hash for key ${key}`));
-    }
+
+    createHash(key, credentials, CACHE_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, model, text, instructions} = {all: true}) {
+  language, voice, deploymentId, engine, text} = {all: true}) {
   logger = logger || noopLogger;
 
   let purgedCount = 0, error;
@@ -33,9 +33,7 @@ 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/synth-audio.js

@@ -3,7 +3,6 @@ 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');
@@ -46,8 +45,7 @@ const {
   JAMBONES_HTTP_PROXY_IP,
   JAMBONES_HTTP_PROXY_PORT,
   JAMBONES_TTS_CACHE_DURATION_MINS,
-  JAMBONES_TTS_TRIM_SILENCE,
-  JAMBONES_AZURE_ENABLE_SSML
+  JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
 } = require('./config');
 const EXPIRES = JAMBONES_TTS_CACHE_DURATION_MINS;
 const OpenAI = require('openai');
@@ -90,15 +88,15 @@ 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 = false, disableTtsStreaming, options, instructions
+  disableTtsCache, renderForCaching = false, disableTtsStreaming, options
 }) {
-  let audioData;
+  let audioBuffer;
   let servedFromCache = false;
   let rtt;
   logger = logger || noopLogger;
 
   assert.ok(['google', 'aws', 'polly', 'microsoft', 'wellsaid', 'nuance', 'nvidia', 'ibm', 'elevenlabs',
-    'whisper', 'deepgram', 'playht', 'rimelabs', 'verbio', 'cartesia', 'inworld', 'resemble'].includes(vendor) ||
+    'whisper', 'deepgram', 'playht', 'rimelabs', 'verbio'].includes(vendor) ||
   vendor.startsWith('custom'),
   `synthAudio supported vendors are google, aws, microsoft, nuance, nvidia and wellsaid ..etc, not ${vendor}`);
   if ('google' === vendor) {
@@ -141,10 +139,6 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
     assert.ok(credentials.api_key, 'synthAudio requires api_key when playht is used');
     assert.ok(credentials.user_id, 'synthAudio requires user_id when playht is used');
     assert.ok(credentials.voice_engine, 'synthAudio requires voice_engine when playht is used');
-  } else if ('inworld' === vendor) {
-    assert.ok(voice, 'synthAudio requires voice when inworld is used');
-    assert.ok(credentials.api_key, 'synthAudio requires api_key when inworld is used');
-    assert.ok(credentials.model_id, 'synthAudio requires model_id when inworld is used');
   } else if ('rimelabs' === vendor) {
     assert.ok(voice, 'synthAudio requires voice when rimelabs is used');
     assert.ok(credentials.api_key, 'synthAudio requires api_key when rimelabs is used');
@@ -163,12 +157,7 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
     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');
-  } else if (vendor === 'resemble') {
-    assert.ok(voice, 'synthAudio requires voice when resemble is used');
-    assert.ok(credentials.api_key, 'synthAudio requires api_key when resemble is used');
+
   }
 
   const key = makeSynthKey({
@@ -177,27 +166,45 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
     language: language || '',
     voice: voice || deploymentId,
     engine,
-    // model or model_id is used to identify the tts cache.
-    model: model || credentials.model_id,
     text,
-    instructions
+    renderForCaching
   });
-
+  let filePath;
+  filePath = makeFilePath({vendor, key, salt, renderForCaching});
   debug(`synth key is ${key}`);
   let cached;
   if (!disableTtsCache) {
     cached = await client.get(key);
+    /**
+    *  If we are using tts streaming and also precaching audio, audio could have been cached by streaming (r8)
+    * or here in speech-utils due to precaching (mp3), so we need to check for both keys.
+    */
+    if (!cached && JAMBONES_EAGERLY_PRE_CACHE_AUDIO) {
+      const preCachekey = makeSynthKey({
+        account_sid,
+        vendor,
+        language: language || '',
+        voice: voice || deploymentId,
+        engine,
+        text,
+        renderForCaching: true
+      });
+      cached = await client.get(preCachekey);
+      if (cached) {
+        // Precache audio is available update filpath with precache file extension.
+        filePath = makeFilePath({vendor, key, salt, renderForCaching: true});
+      }
+    }
   }
   if (cached) {
     // found in cache - extend the expiry and use it
     debug('result WAS found in cache');
     servedFromCache = true;
     stats.increment('tts.cache.requests', ['found:yes']);
-    audioData = JSON.parse(cached);
-    // convert base64 audio to buffer
-    audioData.audioContent = Buffer.from(audioData.audioContent, 'base64');
+    audioBuffer = Buffer.from(cached, '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']);
@@ -205,101 +212,82 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
     const startAt = process.hrtime();
     switch (vendor) {
       case 'google':
-        audioData = await synthGoogle(logger, {credentials, stats, language, voice, gender, text});
+        audioBuffer = await synthGoogle(logger, {credentials, stats, language, voice, gender, text});
         break;
       case 'aws':
       case 'polly':
         vendorLabel = 'aws';
-        audioData = await synthPolly(createHash, retrieveHash, logger,
+        audioBuffer = await synthPolly(createHash, retrieveHash, logger,
           {credentials, stats, language, voice, text, engine});
         break;
       case 'azure':
       case 'microsoft':
         vendorLabel = 'microsoft';
-        audioData = await synthMicrosoft(logger, {credentials, stats, language, voice, text, deploymentId,
-          renderForCaching, disableTtsStreaming});
+        audioBuffer = await synthMicrosoft(logger, {credentials, stats, language, voice, text, deploymentId,
+          filePath, renderForCaching, disableTtsStreaming});
         break;
       case 'nuance':
         model = model || 'enhanced';
-        audioData = await synthNuance(client, logger, {credentials, stats, voice, model, text});
+        audioBuffer = await synthNuance(client, logger, {credentials, stats, voice, model, text});
         break;
       case 'nvidia':
-        audioData = await synthNvidia(client, logger, {credentials, stats, language, voice, model, text,
-          renderForCaching, disableTtsStreaming});
+        audioBuffer = await synthNvidia(client, logger, {credentials, stats, language, voice, model, text});
         break;
       case 'ibm':
-        audioData = await synthIbm(logger, {credentials, stats, voice, text});
+        audioBuffer = await synthIbm(logger, {credentials, stats, voice, text});
         break;
       case 'wellsaid':
-        audioData = await synthWellSaid(logger, {credentials, stats, language, voice, text});
+        audioBuffer = await synthWellSaid(logger, {credentials, stats, language, voice, text, filePath});
         break;
       case 'elevenlabs':
-        audioData = await synthElevenlabs(logger, {
-          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming});
+        audioBuffer = await synthElevenlabs(logger, {
+          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming, filePath
+        });
         break;
       case 'playht':
-        audioData = await synthPlayHT(client, logger, {
-          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 'inworld':
-        audioData = await synthInworld(logger, {
-          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming});
+        audioBuffer = await synthPlayHT(client, logger, {
+          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming, filePath
+        });
         break;
       case 'rimelabs':
-        audioData = await synthRimelabs(logger, {
-          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming});
+        audioBuffer = await synthRimelabs(logger, {
+          credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming, filePath
+        });
         break;
       case 'whisper':
-        audioData = await synthWhisper(logger, {
-          credentials, stats, voice, text, instructions, renderForCaching, disableTtsStreaming});
+        audioBuffer = await synthWhisper(logger, {
+          credentials, stats, voice, text, renderForCaching, disableTtsStreaming});
         break;
       case 'verbio':
-        audioData = await synthVerbio(client, logger, {
+        audioBuffer = await synthVerbio(client, logger, {
           credentials, stats, voice, text, renderForCaching, disableTtsStreaming});
-        if (audioData?.filePath) return audioData;
+        if (audioBuffer?.filePath) return audioBuffer;
         break;
       case 'deepgram':
-        audioData = await synthDeepgram(logger, {credentials, stats, model, text,
+        audioBuffer = await synthDeepgram(logger, {credentials, stats, model, text,
           renderForCaching, disableTtsStreaming});
         break;
-      case 'resemble':
-        audioData = await synthResemble(logger, {
-          credentials, stats, voice, text, options, renderForCaching, disableTtsStreaming});
-        break;
       case vendor.startsWith('custom') ? vendor : 'cant_match_value':
-        audioData = await synthCustomVendor(logger,
-          {credentials, stats, language, voice, text});
+        ({ audioBuffer, filePath } = await synthCustomVendor(logger,
+          {credentials, stats, language, voice, text, filePath}));
         break;
       default:
         assert(`synthAudio: unsupported speech vendor ${vendor}`);
     }
-    if ('filePath' in audioData) return audioData;
+    if ('filePath' in audioBuffer) return audioBuffer;
     const diff = process.hrtime(startAt);
     const time = diff[0] * 1e3 + diff[1] * 1e-6;
     rtt = time.toFixed(0);
     stats.histogram('tts.response_time', rtt, [`vendor:${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, JSON.stringify({
-      ...audioData,
-      audioContent: audioData.audioContent?.toString('base64')
-    }))
+
+    client.setex(key, EXPIRES, audioBuffer.toString('base64'))
       .catch((err) => logger.error(err, `error calling setex on key ${key}`));
   }
 
   return new Promise((resolve, reject) => {
-    const { audioContent, extension } = audioData;
-    const filePath = makeFilePath({
-      key,
-      salt,
-      extension
-    });
-    fs.writeFile(filePath, audioContent, (err) => {
+    fs.writeFile(filePath, audioBuffer, (err) => {
       if (err) return reject(err);
       resolve({filePath, servedFromCache, rtt});
     });
@@ -354,13 +342,7 @@ const synthPolly = async(createHash, retrieveHash, logger,
         .on('data', (chunk) => {
           chunks.push(chunk);
         })
-        .on('end', () => resolve(
-          {
-            audioContent: Buffer.concat(chunks),
-            extension: 'mp3',
-            sampleRate: 8000
-          }
-        ));
+        .on('end', () => resolve(Buffer.concat(chunks)));
     });
   } catch (err) {
     logger.info({err}, 'synthAudio: Error synthesizing speech using aws polly');
@@ -371,48 +353,6 @@ 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}),
@@ -427,11 +367,7 @@ 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 {
-      audioContent: responses[0].audioContent,
-      extension: 'mp3',
-      sampleRate: 8000
-    };
+    return responses[0].audioContent;
   } catch (err) {
     console.error(err);
     logger.info({err, opts}, 'synthAudio: Error synthesizing speech using google');
@@ -462,11 +398,7 @@ const synthIbm = async(logger, {credentials, stats, voice, text}) => {
     for await (const chunk of r.result) {
       chunks.push(chunk);
     }
-    return {
-      audioContent: Buffer.concat(chunks),
-      extension: 'mp3',
-      sampleRate: 8000
-    };
+    return Buffer.concat(chunks);
   } catch (err) {
     logger.info({err, params}, 'synthAudio: Error synthesizing speech using ibm');
     stats.increment('tts.count', ['vendor:ibm', 'accepted:no']);
@@ -476,12 +408,15 @@ 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;
+
   if (use_custom_tts && !content.startsWith('<speak')) {
     /**
      * Note: it seems that to use custom voice ssml is required with the voice attribute
@@ -497,25 +432,17 @@ async function _synthOnPremMicrosoft(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}">${words}</voice></speak>`;
     logger.info({content}, 'synthMicrosoft');
   }
-  else if (JAMBONES_AZURE_ENABLE_SSML && !content.startsWith('<speak')) {
-    // eslint-disable-next-line max-len
-    content = `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="${language}"><voice name="${voice}"><lang xml:lang="${language}">${text}</lang></voice></speak>`;
-  }
 
   try {
-    const trimSilence = JAMBONES_TTS_TRIM_SILENCE;
+    const trimSilence = filePath.endsWith('.r8');
     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 audioContent = await post(custom_tts_endpoint_url, content);
-    return {
-      audioContent,
-      extension: trimSilence ? 'r8' : 'mp3',
-      sampleRate: 8000
-    };
+    const mp3 = await post(custom_tts_endpoint_url, content);
+    return mp3;
   } catch (err) {
     logger.info({err}, '_synthMicrosoftByHttp returned error');
     throw err;
@@ -528,6 +455,7 @@ const synthMicrosoft = async(logger, {
   language,
   voice,
   text,
+  filePath,
   renderForCaching,
   disableTtsStreaming
 }) => {
@@ -537,23 +465,19 @@ const synthMicrosoft = async(logger, {
     let content = text;
     if (use_custom_tts && !content.startsWith('<speak')) {
       /**
-     * Note: it seems that to use custom voice ssml is required with the voice attribute
-     * Otherwise sending plain text we get "Voice does not match"
-     */
+       * Note: it seems that to use custom voice ssml is required with the voice attribute
+       * Otherwise sending plain text we get "Voice does not match"
+       */
       content = `<speak>${text}</speak>`;
     }
 
     if (content.startsWith('<speak>')) {
-    /* microsoft enforces some properties and uses voice xml element so if the user did not supply do it for them */
+      /* microsoft enforces some properties and uses voice xml element so if the user did not supply do it for them */
       const words = content.slice(7, -8).trim().replace(/(\r\n|\n|\r)/gm, ' ');
       // eslint-disable-next-line max-len
       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}">${words}</voice></speak>`;
       logger.info({content}, 'synthMicrosoft');
     }
-    else if (JAMBONES_AZURE_ENABLE_SSML && !content.startsWith('<speak')) {
-      // eslint-disable-next-line max-len
-      content = `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="${language}"><voice name="${voice}"><lang xml:lang="${language}">${text}</lang></voice></speak>`;
-    }
     if (!JAMBONES_DISABLE_TTS_STREAMING && !JAMBONES_DISABLE_AZURE_TTS_STREAMING &&
       !renderForCaching && !disableTtsStreaming) {
       let params = '';
@@ -574,18 +498,17 @@ 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
       });
     }
-    // Azure hosted service
-    const trimSilence = JAMBONES_TTS_TRIM_SILENCE;
+    const trimSilence = filePath.endsWith('.r8');
     const speechConfig = SpeechConfig.fromSubscription(apiKey, region);
     speechConfig.speechSynthesisLanguage = language;
     speechConfig.speechSynthesisVoiceName = voice;
@@ -620,11 +543,7 @@ const synthMicrosoft = async(logger, {
             case ResultReason.SynthesizingAudioCompleted:
               let buffer = Buffer.from(result.audioData);
               if (trimSilence) buffer = trimTrailingSilence(buffer);
-              resolve({
-                audioContent: buffer,
-                extension: trimSilence ? 'r8' : 'mp3',
-                sampleRate: 8000
-              });
+              resolve(buffer);
               synthesizer.close();
               stats.increment('tts.count', ['vendor:microsoft', 'accepted:yes']);
               break;
@@ -654,15 +573,11 @@ const synthWellSaid = async(logger, {credentials, stats, language, voice, gender
       'Accept': 'audio/mpeg',
       'Content-Type': 'application/json'
     });
-    const audioContent = await post('/v1/tts/stream', {
+    const mp3 = await post('/v1/tts/stream', {
       text,
       speaker_id: voice
     });
-    return {
-      audioContent,
-      extension: 'mp3',
-      sampleRate: 8000
-    };
+    return mp3;
   } catch (err) {
     logger.info({err}, 'testWellSaidTts returned error');
     throw err;
@@ -699,8 +614,8 @@ const synthNuance = async(client, logger, {credentials, stats, voice, model, tex
     t.setText(text);
     input.setText(t);
   }
-  const sampleRate = 8000;
-  pcm.setSampleRateHz(sampleRate);
+
+  pcm.setSampleRateHz(8000);
   f.setPcm(pcm);
   p.setAudioFormat(f);
   v.setName(voice);
@@ -724,41 +639,20 @@ const synthNuance = async(client, logger, {credentials, stats, voice, model, tex
         const details = status.getDetails();
         return reject({code, message, details});
       }
-      resolve({
-        audioContent: Buffer.from(response.getAudio()),
-        extension: 'r8',
-        sampleRate
-      });
+      resolve(Buffer.from(response.getAudio()));
     });
   });
 };
 
-const synthNvidia = async(client, logger, {
-  credentials, stats, language,  voice, model, text, renderForCaching, disableTtsStreaming
-}) => {
+const synthNvidia = async(client, logger, {credentials, stats, language,  voice, model, text}) => {
   const {riva_server_uri} = credentials;
-  if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
-    let params = '';
-    params += `{riva_server_uri=${riva_server_uri}`;
-    params += `,voice=${voice}`;
-    params += `,language=${language}`;
-    params += ',write_cache_file=1';
-    params += '}';
-
-    return {
-      filePath: `say:${params}${text.replace(/\n/g, ' ')}`,
-      servedFromCache: false,
-      rtt: 0
-    };
-  }
   let rivaClient, request;
-  const sampleRate = 8000;
   try {
     rivaClient = await createRivaClient(riva_server_uri);
     request = new SynthesizeSpeechRequest();
     request.setVoiceName(voice);
     request.setLanguageCode(language);
-    request.setSampleRateHz(sampleRate);
+    request.setSampleRateHz(8000);
     request.setEncoding(AudioEncoding.LINEAR_PCM);
     request.setText(text);
   } catch (err) {
@@ -772,11 +666,7 @@ const synthNvidia = async(client, logger, {
         logger.info({err, voice, language}, 'error synthesizing speech using Nvidia');
         return reject(err);
       }
-      resolve({
-        audioContent: Buffer.from(response.getAudio()),
-        extension: 'r8',
-        sampleRate
-      });
+      resolve(Buffer.from(response.getAudio()));
     });
   });
 };
@@ -798,13 +688,12 @@ 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 {
-      audioContent: buffer,
-      extension,
-      sampleRate
+      audioBuffer: buffer,
+      filePath: filePath.replace(regex, getFileExtFromMime(mime))
     };
   } catch (err) {
     logger.info({err}, `Vendor ${vendor} returned error`);
@@ -830,13 +719,9 @@ const synthElevenlabs = async(logger, {
     if (opts.voice_settings?.similarity_boost) params += `,similarity_boost=${opts.voice_settings.similarity_boost}`;
     if (opts.voice_settings?.stability) params += `,stability=${opts.voice_settings.stability}`;
     if (opts.voice_settings?.style) params += `,style=${opts.voice_settings.style}`;
-    if (opts.voice_settings?.speed !== null && opts.voice_settings?.speed !== undefined)
-      params += `,speed=${opts.voice_settings.speed}`;
     if (opts.voice_settings?.use_speaker_boost === false) params += ',use_speaker_boost=false';
     if (opts.previous_text) params += `,previous_text=${opts.previous_text}`;
     if (opts.next_text) params += `,next_text=${opts.next_text}`;
-    if (opts.pronunciation_dictionary_locators && Array.isArray(opts.pronunciation_dictionary_locators))
-      params += `,pronunciation_dictionary_locators=${JSON.stringify(opts.pronunciation_dictionary_locators)}`;
     params += '}';
 
     return {
@@ -854,7 +739,7 @@ const synthElevenlabs = async(logger, {
       'Accept': 'audio/mpeg',
       'Content-Type': 'application/json'
     });
-    const audioContent = await post(`/v1/text-to-speech/${voice}${optimize_streaming_latency}`, {
+    const mp3 = await post(`/v1/text-to-speech/${voice}${optimize_streaming_latency}`, {
       text,
       model_id,
       voice_settings: {
@@ -863,11 +748,7 @@ const synthElevenlabs = async(logger, {
       },
       ...opts
     });
-    return {
-      audioContent,
-      extension: 'mp3',
-      sampleRate: 8000
-    };
+    return mp3;
   } catch (err) {
     logger.info({err}, 'synth Elevenlabs returned error');
     stats.increment('tts.count', ['vendor:elevenlabs', 'accepted:no']);
@@ -878,10 +759,11 @@ const synthElevenlabs = async(logger, {
 const synthPlayHT = async(client, logger, {
   credentials, options, stats, voice, language, text, renderForCaching, disableTtsStreaming
 }) => {
-  const {api_key, user_id, voice_engine, playht_tts_uri, options: credOpts} = credentials;
+  const {api_key, user_id, voice_engine, options: credOpts} = credentials;
   const opts = !!options && Object.keys(options).length !== 0 ? options : JSON.parse(credOpts || '{}');
 
-  let synthesizeUrl = playht_tts_uri ? `${playht_tts_uri}/api/v2/tts/stream` : 'https://api.play.ht/api/v2/tts/stream';
+  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 {
@@ -948,7 +830,7 @@ const synthPlayHT = async(client, logger, {
       'Content-Type': 'application/json'
     });
 
-    const audioContent = await post(synthesizeUrl, {
+    const mp3 = await post(synthesizeUrl, {
       text,
       ...(voice_engine === 'Play3.0' && { language }),
       voice,
@@ -957,11 +839,7 @@ const synthPlayHT = async(client, logger, {
       sample_rate: 8000,
       ...opts
     });
-    return {
-      audioContent,
-      extension: 'mp3',
-      sampleRate: 8000
-    };
+    return mp3;
   } catch (err) {
     logger.info({err}, 'synth PlayHT returned error');
     stats.increment('tts.count', ['vendor:playht', 'accepted:no']);
@@ -969,92 +847,22 @@ const synthPlayHT = async(client, logger, {
   }
 };
 
-const synthInworld = async(logger, {
+const synthRimelabs = async(logger, {
   credentials, options, stats, voice, text, renderForCaching, disableTtsStreaming
 }) => {
   const {api_key, model_id, options: credOpts} = credentials;
   const opts = !!options && Object.keys(options).length !== 0 ? options : JSON.parse(credOpts || '{}');
 
-  /* default to using the streaming interface, unless disabled by env var OR we want just a cache file */
-  if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
-    let params = '';
-    params += `{api_key=${api_key}`;
-    params += `,model_id=${model_id}`;
-    params += ',vendor=inworld';
-    params += `,voice=${voice}`;
-    params += ',write_cache_file=1';
-    if (opts.temperature) params += `,temperature=${opts.temperature}`;
-    if (opts.audioConfig?.pitch) params += `,pitch=${opts.pitch}`;
-    if (opts.audioConfig?.speakingRate) params += `,speakingRate=${opts.speakingRate}`;
-    params += '}';
-
-    return {
-      filePath: `say:${params}${text.replace(/\n/g, ' ').replace(/\r/g, ' ')}`,
-      servedFromCache: false,
-      rtt: 0
-    };
-  }
-
-  try {
-    const url = 'https://api.inworld.ai/tts/v1/voice';
-    const sampleRate = 8000;
-    const options = {
-      method: 'POST',
-      headers: {
-        'Authorization': `Basic ${api_key}`,
-        'Content-Type': 'application/json'
-      },
-      body: JSON.stringify({
-        text,
-        voiceId: voice,
-        modelId: model_id,
-        audioConfig: {
-          ...(opts.audioConfig || {}),
-          audioEncoding: 'MP3',
-        }
-      })
-    };
-
-    const response = await fetch(url, options);
-
-    if (!response.ok) {
-      throw new Error(await response.text());
-    }
-    const json = await response.json();
-    return {
-      audioContent: Buffer.from(json.audioContent, 'base64'),
-      extension: 'mp3',
-      sampleRate
-    };
-  } catch (err) {
-    logger.info({err}, 'synth inworld returned error');
-    stats.increment('tts.count', ['vendor:inworld', 'accepted:no']);
-    throw err;
-  }
-};
-
-const synthRimelabs = async(logger, {
-  credentials, options, stats, language, voice, text, renderForCaching, disableTtsStreaming
-}) => {
-  const {api_key, model_id, options: credOpts} = credentials;
-  const opts = !!options && Object.keys(options).length !== 0 ? options : JSON.parse(credOpts || '{}');
-
   /* default to using the streaming interface, unless disabled by env var OR we want just a cache file */
   if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
     let params = '';
     params += `{api_key=${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}`;
     if (opts.reduceLatency) params += `,reduce_latency=${opts.reduceLatency}`;
-    // Arcana model parameters
-    if (opts.temperature) params += `,temperature=${opts.temperature}`;
-    if (opts.repetition_penalty) params += `,repetition_penalty=${opts.repetition_penalty}`;
-    if (opts.top_p) params += `,top_p=${opts.top_p}`;
-    if (opts.max_tokens) params += `,max_tokens=${opts.max_tokens}`;
     params += '}';
 
     return {
@@ -1070,20 +878,14 @@ const synthRimelabs = async(logger, {
       'Accept': 'audio/mp3',
       'Content-Type': 'application/json'
     });
-    const sampleRate = 8000;
-    const audioContent = await post('/v1/rime-tts', {
+    const mp3 = await post('/v1/rime-tts', {
       speaker: voice,
       text,
       modelId: model_id,
-      samplingRate: sampleRate,
-      lang: language,
+      samplingRate: 8000,
       ...opts
     });
-    return {
-      audioContent,
-      extension: 'mp3',
-      sampleRate
-    };
+    return mp3;
   } catch (err) {
     logger.info({err}, 'synth rimelabs returned error');
     stats.increment('tts.count', ['vendor:rimelabs', 'accepted:no']);
@@ -1117,17 +919,13 @@ const synthVerbio = async(client, logger, {credentials, stats, voice, text, rend
       'User-Agent': 'jambonz',
       'Content-Type': 'application/json'
     });
-    const audioContent = await post('/api/v1/synthesize', {
+    const r8 = await post('/api/v1/synthesize', {
       voice_id: voice,
       output_sample_rate: '8k',
       output_encoding: 'pcm16',
       text
     });
-    return {
-      audioContent,
-      extension: 'r8',
-      sampleRate: 8000
-    };
+    return r8;
   } catch (err) {
     logger.info({err}, 'synth Verbio returned error');
     stats.increment('tts.count', ['vendor:verbio', 'accepted:no']);
@@ -1135,8 +933,7 @@ const synthVerbio = async(client, logger, {credentials, stats, voice, text, rend
   }
 };
 
-const synthWhisper = async(logger, {credentials, stats, voice, text, instructions,
-  renderForCaching, disableTtsStreaming}) => {
+const synthWhisper = async(logger, {credentials, stats, voice, text, renderForCaching, disableTtsStreaming}) => {
   const {api_key, model_id, baseURL, timeout, speed} = credentials;
   /* if the env is set to stream then bag out, unless we are specifically rendering to generate a cache file */
   if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
@@ -1147,8 +944,6 @@ const synthWhisper = async(logger, {credentials, stats, voice, text, instruction
     params += `,voice=${voice}`;
     params += ',write_cache_file=1';
     if (speed) params += `,speed=${speed}`;
-    // comma is used to separated parameters in freeswitch tts module
-    if (instructions) params += `,instructions=${instructions.replace(/\n/g, ' ').replace(/,/g, ';')}`;
     params += '}';
 
     return {
@@ -1168,14 +963,9 @@ const synthWhisper = async(logger, {credentials, stats, voice, text, instruction
       model: model_id,
       voice,
       input: text,
-      ...(instructions && {instructions}),
       response_format: 'mp3'
     });
-    return {
-      audioContent: Buffer.from(await mp3.arrayBuffer()),
-      extension: 'mp3',
-      sampleRate: 8000
-    };
+    return Buffer.from(await mp3.arrayBuffer());
   } catch (err) {
     logger.info({err}, 'synth whisper returned error');
     stats.increment('tts.count', ['vendor:openai', 'accepted:no']);
@@ -1207,14 +997,10 @@ const synthDeepgram = async(logger, {credentials, stats, model, text, renderForC
       'Accept': 'audio/mpeg',
       'Content-Type': 'application/json'
     });
-    const audioContent = await post(`/v1/speak?model=${model}`, {
+    const mp3 = await post(`/v1/speak?model=${model}`, {
       text
     });
-    return {
-      audioContent,
-      extension: 'mp3',
-      sampleRate: 8000
-    };
+    return mp3;
   } catch (err) {
     logger.info({err}, 'synth Deepgram returned error');
     stats.increment('tts.count', ['vendor:deepgram', 'accepted:no']);
@@ -1222,151 +1008,26 @@ 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 synthResemble = async(logger, {
-  credentials, options, stats, voice, text, renderForCaching, disableTtsStreaming
-}) => {
-  const {api_key, resemble_tts_uri} = credentials;
-  const {project_uuid, use_hd} = options || {};
-
-  /* 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 = '';
-    params += `{api_key=${api_key}`;
-    params += ',vendor=resemble';
-    params += `,voice=${voice}`;
-    params += ',write_cache_file=1';
-    if (project_uuid) params += `,project_uuid=${project_uuid}`;
-    if (use_hd) params += `,use_hd=${use_hd}`;
-    if (resemble_tts_uri) params += `,endpoint=${resemble_tts_uri}`;
-
-    params += '}';
-
-    return {
-      filePath: `say:${params}${text.replace(/\n/g, ' ').replace(/\r/g, ' ')}`,
-      servedFromCache: false,
-      rtt: 0
-    };
-  }
-
-  try {
-    const baseUrl = resemble_tts_uri || 'https://f.cluster.resemble.ai';
-    const post = bent(baseUrl, 'POST', 'buffer', {
-      'Authorization': `Bearer ${api_key}`,
-      'Content-Type': 'application/json'
-    });
-    const response = await post('/synthesize', {
-      voice_uuid: voice,
-      data: text,
-      sample_rate: 8000,
-      output_format: 'mp3',
-      ...(project_uuid && {project_uuid}),
-      ...(use_hd && {use_hd}),
-    });
-
-    const json = JSON.parse(response.toString('utf8'));
-    const audioContent = Buffer.from(json.audio_content, 'base64');
-    return {
-      audioContent,
-      extension: 'mp3',
-      sampleRate: 8000
-    };
-  } catch (err) {
-    logger.info({err}, 'synth Elevenlabs returned error');
-    stats.increment('tts.count', ['vendor:elevenlabs', 'accepted:no']);
-    throw err;
-  }
-};
-
 const getFileExtFromMime = (mime) => {
   switch (mime) {
     case 'audio/wav':
     case 'audio/x-wav':
-      return ['wav', 8000];
+      return '.wav';
     case /audio\/l16.*rate=8000/.test(mime) ? mime : 'cant match value':
-      return ['r8', 8000];
+      return '.r8';
     case /audio\/l16.*rate=16000/.test(mime) ? mime : 'cant match value':
-      return ['r16', 16000];
+      return '.r16';
     case /audio\/l16.*rate=24000/.test(mime) ? mime : 'cant match value':
-      return ['r24', 24000];
+      return '.r24';
     case /audio\/l16.*rate=32000/.test(mime) ? mime : 'cant match value':
-      return ['r32', 32000];
+      return '.r32';
     case /audio\/l16.*rate=48000/.test(mime) ? mime : 'cant match value':
-      return ['r48', 48000];
+      return '.r48';
     case 'audio/mpeg':
     case 'audio/mp3':
-      return ['mp3', 8000];
+      return '.mp3';
     default:
-      return ['wav', 8000];
+      return '.wav';
   }
 };
 

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 { TMP_FOLDER, HTTP_TIMEOUT } = require('./config');
+const { JAMBONES_DISABLE_TTS_STREAMING, JAMBONES_TTS_TRIM_SILENCE, TMP_FOLDER, HTTP_TIMEOUT } = require('./config');
 
 const debug = require('debug')('jambonz:realtimedb-helpers');
 /**
@@ -17,27 +17,65 @@ const debug = require('debug')('jambonz:realtimedb-helpers');
 //const nuanceClientMap = new Map();
 
 function makeSynthKey({
-  account_sid = '',
-  vendor,
-  language,
-  voice,
-  engine = '',
-  model = '',
-  text,
-  instructions = '',
-}) {
+  account_sid = '', vendor, language, voice, engine = '', text,
+  renderForCaching = false}) {
   const hash = crypto.createHash('sha1');
-  hash.update(`${language}:${vendor}:${voice}:${engine}:${model}:${text}:${instructions}`);
+  hash.update(`${language}:${vendor}:${voice}:${engine}:${text}`);
   const hexHashKey = hash.digest('hex');
   const accountKey = account_sid ? `:${account_sid}` : '';
-  const key = `tts${accountKey}:${hexHashKey}`;
+  const namespace = vendor.startsWith('custom') ? vendor : getFileExtension({vendor, renderForCaching});
+  const key = `tts${accountKey}:${namespace}:${hexHashKey}`;
   return key;
 }
 
-function makeFilePath({key, salt = '', extension}) {
+function makeFilePath({vendor, key, salt = '', renderForCaching = false}) {
+  const extension = getFileExtension({vendor, renderForCaching});
   return `${TMP_FOLDER}/${key.replace('tts:', `tts-${salt}`)}.${extension}`;
 }
 
+function getFileExtension({vendor, renderForCaching = false}) {
+  const mp3Extension = 'mp3';
+  const wavExtension = 'wav';
+  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 'elevenlabs':
+    case 'rimelabs':
+      if (renderForCaching || JAMBONES_DISABLE_TTS_STREAMING) {
+        return mp3Extension;
+      } else {
+        return r8Extension;
+      }
+    case 'playht':
+    case 'deepgram':
+      if (renderForCaching || JAMBONES_DISABLE_TTS_STREAMING) {
+        return mp3Extension;
+      } else {
+        return wavExtension;
+      }
+    case 'nuance':
+    case 'nvidia':
+    case 'verbio':
+      return r8Extension;
+    default:
+      // If vendor is custom
+      if (vendor.startsWith('custom')) {
+        if (renderForCaching || JAMBONES_DISABLE_TTS_STREAMING) {
+          return mp3Extension;
+        } else {
+          return r8Extension;
+        }
+      }
+      return mp3Extension;
+  }
+}
 
 const noopLogger = {
   info: () => {},

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/speech-utils",
-  "version": "0.2.16",
+  "version": "0.1.20",
   "description": "TTS-related speech utilities for jambonz",
   "main": "index.js",
   "author": "Dave Horton",
@@ -26,21 +26,19 @@
   },
   "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",
-    "@cartesia/cartesia-js": "^2.1.0",
-    "@google-cloud/text-to-speech": "^5.5.0",
+    "@google-cloud/text-to-speech": "^5.0.2",
     "@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": "^11.0.0",
+    "ibm-watson": "^8.0.0",
     "microsoft-cognitiveservices-speech-sdk": "1.38.0",
-    "openai": "^4.98.0",
-    "undici": "^7.5.0"
+    "openai": "^4.25.0",
+    "undici": "^6.4.0"
   },
   "devDependencies": {
     "config": "^3.3.11",

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);
@@ -91,17 +91,14 @@ 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);
-    const opts = await synthAudio(stats, {
+    let opts = await synthAudio(stats, {
       vendor: 'google',
       credentials: {
         credentials: {
@@ -112,7 +109,7 @@ GCP_CUSTOM_VOICE_FILE nor GCP_CUSTOM_VOICE_JSON_KEY provided, GCP_CUSTOM_VOICE_M
       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
       }
     });
@@ -124,48 +121,6 @@ GCP_CUSTOM_VOICE_FILE nor GCP_CUSTOM_VOICE_JSON_KEY provided, GCP_CUSTOM_VOICE_M
   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);
@@ -554,7 +509,6 @@ 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');
@@ -562,21 +516,6 @@ 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: {
@@ -686,100 +625,7 @@ 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);
-  }
-  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) => {
+test('rimelabs speech synth tests', async(t) => {
   const fn = require('..');
   const {synthAudio, client} = fn(opts, logger);
 
@@ -789,7 +635,7 @@ test('rimelabs speech synth tests mist', async(t) => {
   }
   const text = 'Hi there and welcome to jambones!';
   try {
-    const opts = await synthAudio(stats, {
+    let opts = await synthAudio(stats, {
       vendor: 'rimelabs',
       credentials: {
         api_key: process.env.RIMELABS_API_KEY,
@@ -799,7 +645,7 @@ test('rimelabs speech synth tests mist', async(t) => {
           reduceLatency: false
         })
       },
-      language: 'eng',
+      language: 'en-US',
       voice: 'amber',
       text,
       renderForCaching: true
@@ -813,40 +659,6 @@ test('rimelabs speech synth tests mist', 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);
@@ -946,12 +758,11 @@ 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, model: i, text: i,
-        instructions: i}), i);
+      await client.set(makeSynthKey({vendor: i, language: i, voice: i, engine: 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`);
 
@@ -966,7 +777,7 @@ 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, instructions: i}), i);
+      await client.set(makeSynthKey({vendor: i, language: i, voice: i, engine: i, text: i}), i);
     }
     // save a specific key to tts cache
     const opts = {vendor: 'aws', language: 'en-US', voice: 'MALE', engine: 'Engine', text: 'Hello World!'};
@@ -982,15 +793,24 @@ 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`);
 
-    process.env.VG_TRIM_TTS_SILENCE = 'true';
+    // 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';    
     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);
@@ -1004,10 +824,10 @@ 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, instructions: i}), i);
+      await client.set(makeSynthKey({account_sid, vendor: i, language: i, voice: i, engine: i, text: 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, instructions: i}), i);
+      await client.set(makeSynthKey({account_sid: account_sid2, vendor: i, language: i, voice: i, engine: i, text: 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}`);