support mod_playht_tts (#304)

* support mod_playht_tts

* wip

* wip

* wip

* wip

* wip

* update speech utils version

lib/routes/api/speech-credentials.js

@@ -6,7 +6,8 @@ const sysError = require('../error');
 const {decrypt, encrypt} = require('../../utils/encrypt-decrypt');
 const {parseAccountSid, parseServiceProviderSid, parseSpeechCredentialSid} = require('./utils');
 const {decryptCredential, testWhisper, testDeepgramTTS,
-  getLanguagesAndVoicesForVendor} = require('../../utils/speech-utils');
+  getLanguagesAndVoicesForVendor,
+  testPlayHT} = require('../../utils/speech-utils');
 const {DbErrorUnprocessableRequest, DbErrorForbidden, DbErrorBadRequest} = require('../../utils/errors');
 const {
   testGoogleTts,
@@ -135,6 +136,8 @@ const encryptCredential = (obj) => {
     auth_token = '',
     cobalt_server_uri,
     model_id,
+    user_id,
+    voice_engine,
     options
   } = obj;
 
@@ -219,6 +222,13 @@ const encryptCredential = (obj) => {
       const elevenlabsData = JSON.stringify({api_key, model_id, options});
       return encrypt(elevenlabsData);
 
+    case 'playht':
+      assert(api_key, 'invalid playht speech credential: api_key is required');
+      assert(user_id, 'invalid playht speech credential: user_id is required');
+      assert(voice_engine, 'invalid voice_engine speech credential: voice_engine is required');
+      const playhtData = JSON.stringify({api_key, user_id, voice_engine, options});
+      return encrypt(playhtData);
+
     case 'assemblyai':
       assert(api_key, 'invalid assemblyai speech credential: api_key is required');
       const assemblyaiData = JSON.stringify({api_key});
@@ -418,6 +428,7 @@ router.put('/:sid', async(req, res) => {
           custom_tts_url,
           cobalt_server_uri,
           model_id,
+          voice_engine,
           options,
           deepgram_stt_uri,
           deepgram_stt_use_tls,
@@ -443,6 +454,7 @@ router.put('/:sid', async(req, res) => {
           custom_tts_url,
           cobalt_server_uri,
           model_id,
+          voice_engine,
           options,
           deepgram_stt_uri,
           deepgram_stt_use_tls,
@@ -724,6 +736,17 @@ router.get('/:sid/test', async(req, res) => {
           SpeechCredential.ttsTestResult(sid, false);
         }
       }
+    } else if (cred.vendor === 'playht') {
+      if (cred.use_for_tts) {
+        try {
+          await testPlayHT(logger, synthAudio, credential);
+          results.tts.status = 'ok';
+          SpeechCredential.ttsTestResult(sid, true);
+        } catch (err) {
+          results.tts = {status: 'fail', reason: err.message};
+          SpeechCredential.ttsTestResult(sid, false);
+        }
+      }
     } else if (cred.vendor === 'assemblyai') {
       const {api_key} = credential;
       if (cred.use_for_stt) {

lib/utils/speech-data/tts-model-playht.js

@@ -0,0 +1,6 @@
+module.exports = [
+  { name: 'PlayHT2.0-turbo', value: 'PlayHT2.0-turbo' },
+  { name: 'PlayHT2.0', value: 'PlayHT2.0' },
+  { name: 'PlayHT1.0', value: 'PlayHT1.0' },
+];
+

lib/utils/speech-utils.js

@@ -21,6 +21,7 @@ const TtsWhisperLanguagesVoices = require('./speech-data/tts-whisper');
 const TtsModelDeepgram = require('./speech-data/tts-model-deepgram');
 const TtsModelElevenLabs = require('./speech-data/tts-model-elevenlabs');
 const TtsModelWhisper = require('./speech-data/tts-model-whisper');
+const TtsModelPlayHT = require('./speech-data/tts-model-playht');
 
 const SttGoogleLanguagesVoices = require('./speech-data/stt-google');
 const SttAwsLanguagesVoices = require('./speech-data/stt-aws');
@@ -240,6 +241,27 @@ const testElevenlabs = async(logger, credentials) => {
   }
 };
 
+const testPlayHT = async(logger, synthAudio, credentials) => {
+  try {
+    await synthAudio(
+      {
+        increment: () => {},
+        histogram: () => {}
+      },
+      {
+        vendor: 'playht',
+        credentials,
+        language: 'en-US',
+        voice: 's3://voice-cloning-zero-shot/d9ff78ba-d016-47f6-b0ef-dd630f59414e/female-cs/manifest.json',
+        text: 'Hi there and welcome to jambones!'
+      }
+    );
+  } catch (err) {
+    logger.info({err}, 'synth Playht returned error');
+    throw err;
+  }
+};
+
 const testWhisper = async(logger, synthAudio, credentials) => {
   try {
     await synthAudio({increment: () => {}, histogram: () => {}},
@@ -428,6 +450,12 @@ function decryptCredential(obj, credential, logger, isObscureKey = true) {
     obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
     obj.model_id = o.model_id;
     obj.options = o.options;
+  } else if ('playht' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+    obj.user_id = o.user_id;
+    obj.voice_engine = o.voice_engine;
+    obj.options = o.options;
   } else if (obj.vendor.startsWith('custom:')) {
     const o = JSON.parse(decrypt(credential));
     obj.auth_token = isObscureKey ? obscureKey(o.auth_token) : o.auth_token;
@@ -488,6 +516,8 @@ async function getLanguagesAndVoicesForVendor(logger, vendor, credential, getTts
       return await getLanguagesVoicesForSoniox(credential, getTtsVoices, logger);
     case 'elevenlabs':
       return await getLanguagesVoicesForElevenlabs(credential, getTtsVoices, logger);
+    case 'playht':
+      return await getLanguagesVoicesForPlayHT(credential, getTtsVoices, logger);
     case 'assemblyai':
       return await getLanguagesVoicesForAssemblyAI(credential, getTtsVoices, logger);
     case 'whisper':
@@ -645,6 +675,49 @@ async function getLanguagesVoicesForElevenlabs(credential) {
   }
 }
 
+const concat = (a) => {
+  return a ? ` ${a},` : '';
+};
+
+async function getLanguagesVoicesForPlayHT(credential) {
+  if (credential) {
+    const get = bent('https://api.play.ht', 'GET', 'json', {
+      'AUTHORIZATION' : credential.api_key,
+      'X-USER-ID': credential.user_id,
+      'Accept': 'application/json'
+    });
+
+    const voices = await get('/api/v2/voices');
+
+    const buildVoice = (d) => {
+      let name = `${d.name} -${concat(d.accent)}${concat(d.age)}${concat(d.gender)}
+${concat(d.loudness)}${concat(d.style)}${concat(d.tempo)}${concat(d.texture)}` ;
+      name = name.endsWith(',') ? name.slice(0, -1) : name;
+      return {
+        value: `${d.id}`,
+        name
+      };
+    };
+
+    const ttsVoices = voices.reduce((acc, voice) => {
+      const languageCode = voice.language_code;
+      const existingLanguage = acc.find((lang) => lang.value === languageCode);
+      if (existingLanguage) {
+        existingLanguage.voices.push(buildVoice(voice));
+      } else {
+        acc.push({
+          value: voice.language_code,
+          name: voice.language,
+          voices: [buildVoice(voice)]
+        });
+      }
+      return acc;
+    }, []);
+    return tranform(ttsVoices, undefined, TtsModelPlayHT);
+  }
+  return tranform(undefined, undefined, TtsModelPlayHT);
+}
+
 async function getLanguagesVoicesForAssemblyAI(credential) {
   return tranform(undefined, SttAssemblyaiLanguagesVoices);
 }
@@ -796,6 +869,7 @@ module.exports = {
   testIbmStt,
   testSonioxStt,
   testElevenlabs,
+  testPlayHT,
   testAssemblyStt,
   testDeepgramTTS,
   getSpeechCredential,