fix azure cannot download mp3 if encodingMp3 = true

* support aws speech by roleArn

* support 3 types of aws  credentials

* wip

* wip

* update speech util version

app.js

@@ -52,6 +52,7 @@ const {
   getTtsVoices,
   getTtsSize,
   purgeTtsCache,
+  getAwsAuthToken,
   synthAudio
 } = require('@jambonz/speech-utils')({}, logger);
 const {
@@ -95,6 +96,7 @@ app.locals = {
   deleteKey,
   getTtsVoices,
   getTtsSize,
+  getAwsAuthToken,
   purgeTtsCache,
   synthAudio,
   lookupAppBySid,

db/jambones-sql.sql

@@ -162,7 +162,7 @@ regex VARCHAR(32) NOT NULL COMMENT 'regex-based pattern match against dialed num
 description VARCHAR(1024),
 priority INTEGER NOT NULL COMMENT 'lower priority routes are attempted first',
 PRIMARY KEY (lcr_route_sid)
-) COMMENT='An ordered list of  digit patterns in an LCR table.  The pat';
+) COMMENT='An ordered list of  digit patterns in an LCR table.  The patterns are tested in sequence until one matches';
 
 CREATE TABLE lcr
 (
@@ -173,7 +173,7 @@ default_carrier_set_entry_sid CHAR(36) COMMENT 'default carrier/route to use whe
 service_provider_sid CHAR(36),
 account_sid CHAR(36),
 PRIMARY KEY (lcr_sid)
-) COMMENT='An LCR (least cost routing) table that is used by a service ';
+) COMMENT='An LCR (least cost routing) table that is used by a service provider or account to make decisions about routing outbound calls when multiple carriers are available.';
 
 CREATE TABLE password_settings
 (
@@ -458,6 +458,7 @@ inbound BOOLEAN NOT NULL COMMENT 'if true, whitelist this IP to allow inbound ca
 outbound BOOLEAN NOT NULL COMMENT 'if true, include in least-cost routing when placing calls to the PSTN',
 voip_carrier_sid CHAR(36) NOT NULL,
 is_active BOOLEAN NOT NULL DEFAULT 1,
+send_options_ping BOOLEAN NOT NULL DEFAULT 0,
 pad_crypto BOOLEAN NOT NULL DEFAULT 0,
 protocol ENUM('udp','tcp','tls', 'tls/srtp') DEFAULT 'udp' COMMENT 'Outbound call protocol',
 PRIMARY KEY (sip_gateway_sid)
@@ -495,7 +496,7 @@ messaging_hook_sid CHAR(36) COMMENT 'webhook to call for inbound SMS/MMS ',
 app_json TEXT,
 speech_synthesis_vendor VARCHAR(64) NOT NULL DEFAULT 'google',
 speech_synthesis_language VARCHAR(12) NOT NULL DEFAULT 'en-US',
-speech_synthesis_voice VARCHAR(64),
+speech_synthesis_voice VARCHAR(256),
 speech_synthesis_label VARCHAR(64),
 speech_recognizer_vendor VARCHAR(64) NOT NULL DEFAULT 'google',
 speech_recognizer_language VARCHAR(64) NOT NULL DEFAULT 'en-US',
@@ -503,7 +504,7 @@ speech_recognizer_label VARCHAR(64),
 use_for_fallback_speech BOOLEAN DEFAULT false,
 fallback_speech_synthesis_vendor VARCHAR(64),
 fallback_speech_synthesis_language VARCHAR(12),
-fallback_speech_synthesis_voice VARCHAR(64),
+fallback_speech_synthesis_voice VARCHAR(256),
 fallback_speech_synthesis_label VARCHAR(64),
 fallback_speech_recognizer_vendor VARCHAR(64),
 fallback_speech_recognizer_language VARCHAR(64),

db/jambones.sqs

@@ -551,7 +551,7 @@
         </location>
         <size>
             <width>293.00</width>
-            <height>540.00</height>
+            <height>560.00</height>
         </size>
         <zorder>6</zorder>
         <SQLField>
@@ -2332,7 +2332,7 @@
         </location>
         <size>
             <width>281.00</width>
-            <height>240.00</height>
+            <height>260.00</height>
         </size>
         <zorder>7</zorder>
         <SQLField>
@@ -2399,10 +2399,18 @@
             <notNull><![CDATA[1]]></notNull>
             <uid><![CDATA[27D4A5BD-8093-4ADD-B5B5-D546844206F9]]></uid>
         </SQLField>
+        <SQLField>
+            <name><![CDATA[send_options_ping]]></name>
+            <type><![CDATA[BOOLEAN]]></type>
+            <defaultValue><![CDATA[0]]></defaultValue>
+            <notNull><![CDATA[1]]></notNull>
+            <uid><![CDATA[E04C19A2-12BF-443F-AB61-96990224A18D]]></uid>
+        </SQLField>
         <SQLField>
             <name><![CDATA[pad_crypto]]></name>
             <type><![CDATA[BOOLEAN]]></type>
             <defaultValue><![CDATA[0]]></defaultValue>
+            <forcedUnique><![CDATA[0]]></forcedUnique>
             <notNull><![CDATA[1]]></notNull>
             <uid><![CDATA[C5C0043B-100A-4476-BF01-BE0777AE27C0]]></uid>
         </SQLField>
@@ -2560,7 +2568,7 @@
         </SQLField>
         <SQLField>
             <name><![CDATA[speech_synthesis_voice]]></name>
-            <type><![CDATA[VARCHAR(64)]]></type>
+            <type><![CDATA[VARCHAR(256)]]></type>
             <notNull><![CDATA[0]]></notNull>
             <uid><![CDATA[929D66F0-64B9-4D7C-AB4B-24F131E1178F]]></uid>
         </SQLField>
@@ -2610,7 +2618,7 @@
         </SQLField>
         <SQLField>
             <name><![CDATA[fallback_speech_synthesis_voice]]></name>
-            <type><![CDATA[VARCHAR(64)]]></type>
+            <type><![CDATA[VARCHAR(256)]]></type>
             <notNull><![CDATA[0]]></notNull>
             <uid><![CDATA[6A0E92C9-32B9-4179-A893-3DADF5DD7728]]></uid>
         </SQLField>
@@ -3097,17 +3105,17 @@
         <overviewPanelHidden><![CDATA[0]]></overviewPanelHidden>
         <pageBoundariesVisible><![CDATA[0]]></pageBoundariesVisible>
         <PageGridVisible><![CDATA[0]]></PageGridVisible>
-        <RightSidebarWidth><![CDATA[1924.000000]]></RightSidebarWidth>
+        <RightSidebarWidth><![CDATA[1235.000000]]></RightSidebarWidth>
         <sidebarIndex><![CDATA[2]]></sidebarIndex>
         <snapToGrid><![CDATA[0]]></snapToGrid>
-        <SourceSidebarWidth><![CDATA[0.000000]]></SourceSidebarWidth>
+        <SourceSidebarWidth><![CDATA[312.000000]]></SourceSidebarWidth>
         <SQLEditorFileFormatVersion><![CDATA[4]]></SQLEditorFileFormatVersion>
         <uid><![CDATA[58C99A00-06C9-478C-A667-C63842E088F3]]></uid>
-        <windowHeight><![CDATA[985.000000]]></windowHeight>
-        <windowLocationX><![CDATA[-1307.000000]]></windowLocationX>
-        <windowLocationY><![CDATA[1008.000000]]></windowLocationY>
-        <windowScrollOrigin><![CDATA[{1.5, 786}]]></windowScrollOrigin>
-        <windowWidth><![CDATA[2201.000000]]></windowWidth>
+        <windowHeight><![CDATA[870.000000]]></windowHeight>
+        <windowLocationX><![CDATA[-1164.000000]]></windowLocationX>
+        <windowLocationY><![CDATA[1131.000000]]></windowLocationY>
+        <windowScrollOrigin><![CDATA[{0.5, 0}]]></windowScrollOrigin>
+        <windowWidth><![CDATA[1512.000000]]></windowWidth>
     </SQLDocumentInfo>
     <AllowsIndexRenamingOnInsert><![CDATA[1]]></AllowsIndexRenamingOnInsert>
     <defaultLabelExpanded><![CDATA[1]]></defaultLabelExpanded>

db/upgrade-jambonz-db.js

@@ -190,7 +190,12 @@ const sql = {
     'ALTER TABLE google_custom_voices ADD FOREIGN KEY speech_credential_sid_idxfk (speech_credential_sid) REFERENCES speech_credentials (speech_credential_sid) ON DELETE CASCADE',
     'ALTER TABLE clients ADD COLUMN allow_direct_queue_calling BOOLEAN NOT NULL DEFAULT 1',
     'ALTER TABLE clients ADD COLUMN allow_direct_user_calling BOOLEAN NOT NULL DEFAULT 1',
-    'ALTER TABLE clients ADD COLUMN allow_direct_app_calling BOOLEAN NOT NULL DEFAULT 1'
+    'ALTER TABLE clients ADD COLUMN allow_direct_app_calling BOOLEAN NOT NULL DEFAULT 1',
+  ],
+  9000: [
+    'ALTER TABLE sip_gateways ADD COLUMN send_options_ping BOOLEAN NOT NULL DEFAULT 0',
+    'ALTER TABLE applications MODIFY COLUMN speech_synthesis_voice VARCHAR(256)',
+    'ALTER TABLE applications MODIFY COLUMN fallback_speech_synthesis_voice VARCHAR(256)',
   ]
 };
 
@@ -223,6 +228,7 @@ const doIt = async() => {
         if (val < 8003) upgrades.push(...sql['8003']);
         if (val < 8004) upgrades.push(...sql['8004']);
         if (val < 8005) upgrades.push(...sql['8005']);
+        if (val < 9000) upgrades.push(...sql['9000']);
 
         // perform all upgrades
         logger.info({upgrades}, 'applying schema upgrades..');

lib/models/voip-carrier.js

@@ -61,6 +61,10 @@ VoipCarrier.fields = [
     name: 'requires_register',
     type: 'number'
   },
+  {
+    name: 'register_use_tls',
+    type: 'number'
+  },
   {
     name: 'register_username',
     type: 'string'

lib/record/s3-multipart-upload-stream.js

@@ -16,7 +16,7 @@ class S3MultipartUploadStream extends Writable {
     this.partNumber = 1;
     this.multipartETags = [];
     this.buffer = Buffer.alloc(0);
-    this.minPartSize = 2 * 1024 * 1024; // 5 MB
+    this.minPartSize = 5 * 1024 * 1024; // 5 MB
     this.s3 = new S3Client(opts.bucketCredential);
     this.metadata = opts.metadata;
   }

lib/record/upload.js

@@ -3,6 +3,7 @@ const Websocket = require('ws');
 const PCMToMP3Encoder = require('./encoder');
 const wav = require('wav');
 const { getUploader } = require('./utils');
+const { pipeline } = require('stream');
 
 async function upload(logger, socket) {
   socket._recvInitialMetadata = false;
@@ -60,22 +61,19 @@ async function upload(logger, socket) {
               bitrate: 128
             }, logger);
           }
-          const handleError = (err, streamType) => {
-            logger.error(
-              { err },
-              `Error while streaming for vendor: ${obj.vendor}, pipe: ${streamType}: ${err.message}`
-            );
-          };
 
           /* start streaming data */
-          const duplex = Websocket.createWebSocketStream(socket);
-          duplex
-            .on('error', (err) => handleError(err, 'duplex'))
-            .pipe(encoder)
-            .on('error', (err) => handleError(err, 'encoder'))
-            .pipe(uploadStream)
-            .on('error', (err) => handleError(err, 'uploadStream'));
-
+          pipeline(
+            Websocket.createWebSocketStream(socket),
+            encoder,
+            uploadStream,
+            (error) => {
+              if (error) {
+                logger.error({ error }, 'pipeline error, cannot upload data to storage');
+                socket.close();
+              }
+            }
+          );
         } else {
           logger.info(`account ${accountSid} does not have any bucket credential, close the socket`);
           socket.close();

lib/routes/api/accounts.js

@@ -265,7 +265,8 @@ function validateUpdateCall(opts) {
     'sip_request',
     'record',
     'tag',
-    'dtmf'
+    'dtmf',
+    'conferenceParticipantAction'
   ]
     .reduce((acc, prop) => (opts[prop] ? ++acc : acc), 0);
 
@@ -316,6 +317,19 @@ function validateUpdateCall(opts) {
   if (opts.tag && (typeof opts.tag !== 'object' || Array.isArray(opts.tag) || opts.tag === null)) {
     throw new DbErrorBadRequest('invalid tag data');
   }
+  if (opts.conferenceParticipantAction) {
+    if (!['tag', 'untag', 'coach', 'uncoach', 'mute', 'unmute', 'hold', 'unhold']
+      .includes(opts.conferenceParticipantAction.action)) {
+      throw new DbErrorBadRequest(
+        `conferenceParticipantAction invalid action property ${opts.conferenceParticipantAction.action}`);
+    }
+    if ('tag' == opts.conferenceParticipantAction.action && !opts.tag) {
+      throw new DbErrorBadRequest('conferenceParticipantAction requires tag property when action is \'tag\'');
+    }
+    if ('coach' == opts.conferenceParticipantAction.action && !opts.tag) {
+      throw new DbErrorBadRequest('conferenceParticipantAction requires tag property when action is \'coach\'');
+    }
+  }
 }
 
 function validateTo(to) {

lib/routes/api/sip-realm.js

@@ -31,6 +31,7 @@ router.post('/:sip_realm', async(req, res) => {
       const [sbcs] = await promisePool.query('SELECT ipv4 from sbc_addresses');
       if (sbcs.length === 0) throw new Error('no SBC addresses provisioned in the database!');
       const ips = sbcs.map((s) => s.ipv4);
+      const uniqueIps = [...new Set(ips)];
 
       /* retrieve existing dns records */
       const [old_recs] = await promisePool.query('SELECT record_id from dns_records WHERE account_sid = ?',
@@ -48,7 +49,7 @@ router.post('/:sip_realm', async(req, res) => {
       }
 
       /* add the dns records */
-      const records = await createDnsRecords(logger, domain, subdomain, ips);
+      const records = await createDnsRecords(logger, domain, subdomain, uniqueIps);
       if (!records) throw new Error(`failure updating dns records for ${sip_realm}`);
       const values = records.map((r) => {
         return `('${uuid()}', '${account_sid}', '${r.type}', ${r.id})`;

lib/routes/api/speech-credentials.js

@@ -6,7 +6,9 @@ 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,
+  testRimelabs} = require('../../utils/speech-utils');
 const {DbErrorUnprocessableRequest, DbErrorForbidden, DbErrorBadRequest} = require('../../utils/errors');
 const {
   testGoogleTts,
@@ -111,6 +113,7 @@ const encryptCredential = (obj) => {
     secret_access_key,
     aws_region,
     api_key,
+    role_arn,
     region,
     client_id,
     secret,
@@ -135,8 +138,9 @@ const encryptCredential = (obj) => {
     auth_token = '',
     cobalt_server_uri,
     model_id,
-    options,
-    use_streaming
+    user_id,
+    voice_engine,
+    options
   } = obj;
 
   switch (vendor) {
@@ -152,10 +156,17 @@ const encryptCredential = (obj) => {
       return encrypt(service_key);
 
     case 'aws':
-      assert(access_key_id, 'invalid aws speech credential: access_key_id is required');
-      assert(secret_access_key, 'invalid aws speech credential: secret_access_key is required');
-      assert(aws_region, 'invalid aws speech credential: aws_region is required');
-      const awsData = JSON.stringify({aws_region, access_key_id, secret_access_key});
+      // AWS polly can work for 3 types of credentials:
+      // 1/ access_key_id and secret_access_key
+      // 2/ RoleArn Assume role
+      // 3/ RoleArn assigned to instance profile where will run this application
+      const awsData = JSON.stringify(
+        {
+          aws_region,
+          ...(access_key_id && {access_key_id}),
+          ...(secret_access_key && {secret_access_key}),
+          ...(role_arn && {role_arn}),
+        });
       return encrypt(awsData);
 
     case 'microsoft':
@@ -217,9 +228,22 @@ const encryptCredential = (obj) => {
     case 'elevenlabs':
       assert(api_key, 'invalid elevenLabs speech credential: api_key is required');
       assert(model_id, 'invalid elevenLabs speech credential: model_id is required');
-      const elevenlabsData = JSON.stringify({api_key, model_id, options, use_streaming});
+      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 'rimelabs':
+      assert(api_key, 'invalid rimelabs speech credential: api_key is required');
+      assert(model_id, 'invalid rimelabs speech credential: model_id is required');
+      const rimelabsData = JSON.stringify({api_key, model_id, options});
+      return encrypt(rimelabsData);
+
     case 'assemblyai':
       assert(api_key, 'invalid assemblyai speech credential: api_key is required');
       const assemblyaiData = JSON.stringify({api_key});
@@ -419,8 +443,8 @@ router.put('/:sid', async(req, res) => {
           custom_tts_url,
           cobalt_server_uri,
           model_id,
+          voice_engine,
           options,
-          use_streaming,
           deepgram_stt_uri,
           deepgram_stt_use_tls,
         } = req.body;
@@ -445,8 +469,8 @@ router.put('/:sid', async(req, res) => {
           custom_tts_url,
           cobalt_server_uri,
           model_id,
+          voice_engine,
           options,
-          use_streaming,
           deepgram_stt_uri,
           deepgram_stt_use_tls,
         };
@@ -525,12 +549,13 @@ router.get('/:sid/test', async(req, res) => {
       }
     }
     else if (cred.vendor === 'aws') {
+      const {getTtsVoices, getAwsAuthToken} = req.app.locals;
       if (cred.use_for_tts) {
-        const {getTtsVoices} = req.app.locals;
         try {
           await testAwsTts(logger, getTtsVoices, {
             accessKeyId: credential.access_key_id,
             secretAccessKey: credential.secret_access_key,
+            roleArn: credential.role_arn,
             region: credential.aws_region || process.env.AWS_REGION
           });
           results.tts.status = 'ok';
@@ -542,9 +567,10 @@ router.get('/:sid/test', async(req, res) => {
       }
       if (cred.use_for_stt) {
         try {
-          await testAwsStt(logger, {
+          await testAwsStt(logger, getAwsAuthToken, {
             accessKeyId: credential.access_key_id,
             secretAccessKey: credential.secret_access_key,
+            roleArn: credential.role_arn,
             region: credential.aws_region || process.env.AWS_REGION
           });
           results.stt.status = 'ok';
@@ -727,6 +753,33 @@ 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) {
+          let reason = err.message;
+          // if error is from bent, let get the body
+          try {
+            reason = await err.text();
+          } catch {}
+          results.tts = {status: 'fail', reason};
+          SpeechCredential.ttsTestResult(sid, false);
+        }
+      }
+    } else if (cred.vendor === 'rimelabs') {
+      if (cred.use_for_tts) {
+        try {
+          await testRimelabs(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/routes/api/tts-cache.js

@@ -10,6 +10,7 @@ const Account = require('../../models/account');
 const sysError = require('../error');
 const { getSpeechCredential, decryptCredential } = require('../../utils/speech-utils');
 const PCMToMP3Encoder = require('../../record/encoder');
+const { pipeline } = require('stream');
 
 router.delete('/', async(req, res) => {
   const {purgeTtsCache} = req.app.locals;
@@ -69,6 +70,8 @@ router.post('/Synthesize', async(req, res) => {
         voice = arr[1];
         model = arr[2];
       }
+    }  else if (cred.vendor === 'deepgram') {
+      model = voice;
     }
     const stats = {
       histogram: () => {},
@@ -84,7 +87,8 @@ router.post('/Synthesize', async(req, res) => {
       model,
       salt,
       credentials: cred,
-      disableTtsCache: false
+      disableTtsCache: false,
+      disableTtsStreaming: true
     });
 
     let contentType = 'audio/mpeg';
@@ -92,17 +96,27 @@ router.post('/Synthesize', async(req, res) => {
     let readStream = fs.createReadStream(filePath);
     if (['nuance', 'nvidia'].includes(cred.vendor) ||
       (
-        process.env.JAMBONES_TTS_TRIM_SILENCE &&
+        (process.env.JAMBONES_TTS_TRIM_SILENCE || !process.env.JAMBONES_DISABLE_TTS_STREAMING) &&
         ['microsoft', 'azure'].includes(cred.vendor)
       )
     ) {
       if (encodingMp3) {
-        readStream = readStream
-          .pipe(new PCMToMP3Encoder({
+        readStream = pipeline(
+          readStream,
+          new PCMToMP3Encoder({
             channels: 1,
             sampleRate: 8000,
             bitRate: 128
-          }, logger));
+          }, logger),
+          (err) => {
+            if (err) {
+              logger.error('ttscache/Synthesize failed:', err);
+              if (!res.headersSent) {
+                res.status(500).end('Server error');
+              }
+            }
+          }
+        );
       } else {
         contentType = 'application/octet-stream';
       }
@@ -110,10 +124,17 @@ router.post('/Synthesize', async(req, res) => {
     res.writeHead(200, {
       'Content-Type': contentType,
     });
-    readStream.pipe(res);
-    readStream.on('end', () => {
-      fs.unlink(filePath, (err) => {
-        if (err) throw err;
+
+    pipeline(readStream, res, (err) => {
+      if (err) {
+        logger.error('ttscache/Synthesize failed:', err);
+        if (!res.headersSent) {
+          res.status(500).end('Server error');
+        }
+      }
+
+      fs.unlink(filePath, (unlinkErr) => {
+        if (unlinkErr) throw unlinkErr;
         logger.info(`${filePath} was deleted`);
       });
     });

lib/swagger/swagger.yaml

@@ -1108,6 +1108,9 @@ paths:
                 requires_register:
                   type: boolean
                   description: wehther this provider requires us to send a REGISTER to them in order to receive calls
+                register_use_tls:
+                  type: boolean
+                  description: wehther this provider requires us to send a REGISTER use TLS protocol
                 register_username:
                   type: string
                   description: sip username to authenticate with, if registration is required
@@ -4149,6 +4152,22 @@ paths:
                       type: string 
                     siprecServerURL:
                       type: string
+                conferenceParticipantAction:
+                  type: object
+                  properties:
+                    action:
+                      type: string
+                      enum:
+                        - tag
+                        - untag
+                        - coach
+                        - uncoach
+                        - mute
+                        - unmute
+                        - hold
+                        - unhold
+                    tag:
+                      type: string
       responses:
         200:
           description: Accepted

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

@@ -1,9 +1,14 @@
 module.exports = [
-  { name: 'Aurora English (US) Female', value: 'alpha-aurora-en-v2' },
-  { name: 'Asteria English (US) Female', value: 'alpha-asteria-en-v2' },
-  { name: 'Artemis English (UK) Female', value: 'alpha-artemis-en-v3' },
-  { name: 'Andromeda English (US) Female', value: 'alpha-andromeda-en-v3' },
-  { name: 'Stella	English (UK) Female', value: 'alpha-stella-en-v2' },
-  { name: 'Orion	English (US)	Male', value: 'alpha-orion-en-v2' },
-  { name: 'Atlas	English (US)	Male', value: 'alpha-atlas-en-v3' },
+  { name: 'Asteria English (US) Female', value: 'aura-asteria-en' },
+  { name: 'Luna English (US) Female', value: 'aura-luna-en' },
+  { name: 'Stella English (US) Female', value: 'aura-stella-en' },
+  { name: 'Stella English (UK) Female', value: 'aura-athena-en' },
+  { name: 'Hera English (US) Female', value: 'aura-hera-en' },
+  { name: 'Orion English (US) Male', value: 'aura-orion-en' },
+  { name: 'Arcas English (US) Male', value: 'aura-arcas-en' },
+  { name: 'Perseus English (US) Male', value: 'aura-perseus-en' },
+  { name: 'Angus English (Ireland) Male', value: 'aura-angus-en' },
+  { name: 'Orpheus English (US) Male', value: 'aura-orpheus-en' },
+  { name: 'Helios English (UK) Male', value: 'aura-helios-en' },
+  { name: 'Zeus English (US) Male', value: 'aura-zeus-en' },
 ];

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-data/tts-model-rimelabs.js

@@ -0,0 +1,5 @@
+module.exports = [
+  { name: 'Mist', value: 'mist' },
+  { name: 'V1', value: 'v1' },
+];
+

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

@@ -0,0 +1,710 @@
+module.exports = [
+  {
+    value: 'en-US',
+    name: 'English (US)',
+    voices: [
+      {
+        value:
+          's3://mockingbird-prod/abigail_vo_6661b91f-4012-44e3-ad12-589fbdee9948/voices/speaker/manifest.json',
+        name: 'Abigail - american, female, narrative, smooth',
+      },
+      {
+        value: 'abram',
+        name: 'Abram - british, old, male, low, narrative, slow, round',
+      },
+      {
+        value: 'adolfo',
+        name: 'Adolfo - american, adult, male, neutral, narrative, fast, thick',
+      },
+      {
+        value: 'adrian',
+        name: 'Adrian - american, old, male, neutral, narrative, fast, thick',
+      },
+      {
+        value: 'ahmed',
+        name: 'Logan - british, old, male, neutral, narrative, neutral, thick',
+      },
+      {
+        value: 'alex',
+        name: 'Alex - british, adult, male, high, narrative, slow, thick',
+      },
+      {
+        value: 'alexander',
+        name: 'Alexander - british, old, male, high, narrative, fast, thick',
+      },
+      {
+        value: 'alfonso',
+        name: 'Alfonso - american, adult, male, neutral, videos, neutral, gravelly',
+      },
+      {
+        value: 'alphonso',
+        name: 'Alphonso - american, adult, female, low, videos, neutral, smooth',
+      },
+      {
+        value: 'amado',
+        name: 'Amado - american, old, male, low, narrative, fast, smooth',
+      },
+      {
+        value: 'anny',
+        name: 'Anny - american, youth, female, neutral, narrative, neutral, thick',
+      },
+      {
+        value: 'anthony',
+        name: 'Anthony - american, adult, male, neutral, training, slow, thick',
+      },
+      {
+        value: 'spencer',
+        name: 'April - british, adult, female, neutral, narrative, slow, smooth',
+      },
+      {
+        value: 'victor',
+        name: 'Ariana - american, youth, female, high, videos, fast, thick',
+      },
+      {
+        value: 'arthur',
+        name: 'Arthur - british, adult, male, neutral, narrative, neutral, smooth',
+      },
+      {
+        value: 'aubrey',
+        name: 'Aubrey - british, adult, male, neutral, videos, neutral, smooth',
+      },
+      {
+        value: 'hipolito',
+        name: 'Audrey - american, adult, female, low, narrative, slow, round',
+      },
+      {
+        value: 'aurora',
+        name: 'Aurora - british, adult, female, low, training, slow, round',
+      },
+      {
+        value: 'axel',
+        name: 'Axel - american, adult, male, neutral, narrative, fast, thick',
+      },
+      {
+        value:
+          's3://mockingbird-prod/ayla_vo_commercials_d66900d5-69f5-476f-9bd6-8eab2936dda3/voices/speaker/manifest.json',
+        name: 'Ayla (Advertising) - american, female, advertising',
+      },
+      {
+        value:
+          's3://mockingbird-prod/ayla_vo_expressive_16095e08-b9e8-429b-947c-47a75e41053b/voices/speaker/manifest.json',
+        name: 'Ayla (Expressive) - american, female, narrative',
+      },
+      {
+        value:
+          's3://mockingbird-prod/ayla_vo_meditation_d11dd9da-b5f1-4709-95a6-e6d5dc77614a/voices/speaker/manifest.json',
+        name: 'Ayla (Meditation) - american, female, meditation',
+      },
+      {
+        value:
+          's3://mockingbird-prod/ayla_vo_narrative_d8199dfd-b50f-40c7-9d99-e203ba5f4152/voices/speaker/manifest.json',
+        name: 'Ayla (Narrative) - american, female, narrative',
+      },
+      {
+        value:
+          's3://mockingbird-prod/ayla_vo_training_e6751ca5-e47c-4c4b-ad05-d3a194417600/voices/speaker/manifest.json',
+        name: 'Ayla (Training) - american, female, training',
+      },
+      {
+        value: 'benton',
+        name: 'Benton - american, old, male, high, videos, fast, smooth',
+      },
+      {
+        value: 'bertram',
+        name: 'Bertram - british, adult, male, low, narrative, neutral, gravelly',
+      },
+      {
+        value: 'bill',
+        name: 'Harper - american, adult, female, high, videos, fast, smooth',
+      },
+      {
+        // eslint-disable-next-line max-len
+        value:'s3://mockingbird-prod/nathan_drake_carmelo_pampillonio_7d540ad6-7d32-41f6-8d53-2584901aa03d/voices/speaker/manifest.json',
+        name: 'Billy - american, male, gaming',
+      },
+      {
+        value: 'blaine',
+        name: 'Blaine - british, adult, male, high, narrative, neutral, thick',
+      },
+      {
+        value: 'booker',
+        name: 'Booker - british, youth, male, neutral, narrative, neutral, round',
+      },
+      {
+        value: 'bret',
+        name: 'Bret - american, adult, female, neutral, narrative, slow, smooth',
+      },
+      {
+        value: 'bruce',
+        name: 'Bruce - british, adult, male, high, training, fast, thick',
+      },
+      {
+        value: 'bryan',
+        name: 'Bryan - american, adult, male, low, videos, fast, gravelly',
+      },
+      {
+        value: 'carlo',
+        name: 'Carlo - british, adult, male, neutral, advertising, neutral, smooth',
+      },
+      {
+        value: 'carter',
+        name: 'Carter - american, adult, male, neutral, narrative, neutral, thick',
+      },
+      {
+        value: 'charles',
+        name: 'Charles - american, adult, male, neutral, narrative, neutral, round',
+      },
+      {
+        value: 'charlotte',
+        name: 'Charlotte - canadian, adult, female, low, narrative, neutral, smooth',
+      },
+      {
+        value:
+          's3://voice-cloning-zero-shot/028a32d4-6a79-4ca3-a303-d6559843114b/chris/manifest.json',
+        name: 'Chris - american, adult, male,',
+      },
+      {
+        value: 'chuck',
+        name: 'Chuck - british, adult, male, neutral, videos, slow, round',
+      },
+      {
+        value: 'clark',
+        name: 'Clark - british, old, male, neutral, narrative, slow, smooth',
+      },
+      {
+        value: 'clifton',
+        name: 'Clifton - american, old, male, high, narrative, neutral, gravelly',
+      },
+      {
+        value: 'hayden',
+        name: 'Cooper - american, adult, male, neutral, narrative, neutral, round',
+      },
+      {
+        value: 'daisy',
+        name: 'Daisy - british, adult, female, low, narrative, neutral, gravelly',
+      },
+      {
+        value: 'dane',
+        name: 'Dane - american, adult, male, neutral, videos, neutral, round',
+      },
+      {
+        value: 'daniel',
+        name: 'Daniel - canadian, adult, male, low, narrative, neutral, smooth',
+      },
+      {
+        value: 'darnell',
+        name: 'Darnell - american, youth, male, neutral, narrative, neutral, smooth',
+      },
+      {
+        value: 'daron',
+        name: 'Daron - american, old, male, low, narrative, slow, round',
+      },
+      {
+        value: 'darrell',
+        name: 'Darrell - british, adult, male, neutral, advertising, neutral, thick',
+      },
+      {
+        value: 's3://peregrine-voices/a10/manifest.json',
+        name: 'Davis - american, adult, male,',
+      },
+      {
+        value: 'ignacio',
+        name: 'Delilah - american, adult, female, neutral, narrative, slow, smooth',
+      },
+      {
+        value: 'denis',
+        name: 'Eleanor - british, adult, female, neutral, advertising, neutral, smooth',
+      },
+      {
+        value: 'dick',
+        name: 'Dick - american, adult, male, neutral, training, fast, smooth',
+      },
+      {
+        value: 'domenic',
+        name: 'Domenic - british, adult, male, high, videos, neutral, thick',
+      },
+      {
+        value: 's3://peregrine-voices/donna_meditation_saad/manifest.json',
+        name: 'Donna (Meditation) - american, female, meditation',
+      },
+      {
+        value: 's3://peregrine-voices/donna_parrot_saad/manifest.json',
+        name: 'Donna (Narrative) - american, female, narrative',
+      },
+      {
+        value: 'donovan',
+        name: 'Donovan - american, adult, male, low, narrative, neutral, smooth',
+      },
+      {
+        value: 'dudley',
+        name: 'Dudley - american, old, male, low, narrative, fast, smooth',
+      },
+      {
+        value: 'dylan',
+        name: 'Dylan - british, old, male, high, gaming, slow, smooth',
+      },
+      {
+        value: 'earle',
+        name: 'Earle - british, adult, male, high, narrative, neutral, gravelly',
+      },
+      {
+        value: 'efren',
+        name: 'Efren - american, adult, male, neutral, training, slow, thick',
+      },
+      {
+        value: 'denis',
+        name: 'Eleanor - british, adult, female, neutral, advertising, neutral, smooth',
+      },
+      {
+        value: 'elijah',
+        name: 'Elijah - american, old, male, neutral, training, neutral, gravelly',
+      },
+      {
+        value: 'ellie',
+        name: 'Ellie - american, adult, female, low, training, slow, smooth',
+      },
+      {
+        value: 'erasmo',
+        name: 'Erasmo - american, old, male, low, training, fast, smooth',
+      },
+      {
+        value: 's3://peregrine-voices/evelyn 2 saad parrot/manifest.json',
+        name: 'Evelyn - american, adult, female, low, videos, neutral, smooth',
+      },
+      {
+        value: 'fletcher',
+        name: 'Fletcher - british, adult, male, neutral, narrative, fast, gravelly',
+      },
+      {
+        value: 'florencio',
+        name: 'Madison - british, old, female, neutral, narrative, slow, round',
+      },
+      {
+        value: 'flynn',
+        name: 'Flynn - british, adult, male, neutral, narrative, fast, round',
+      },
+      {
+        value: 'gabriel',
+        name: 'Samantha - american, old, female, neutral, narrative, neutral, thick',
+      },
+      {
+        value: 'greg',
+        name: 'Greg - british, adult, male, high, narrative, slow, round',
+      },
+      {
+        value: 'harold',
+        name: 'Harold - american, adult, male, neutral, narrative, slow, smooth',
+      },
+      {
+        value: 'bill',
+        name: 'Harper - american, adult, female, high, videos, fast, smooth',
+      },
+      {
+        value: 'harris',
+        name: 'Harris - british, adult, male, low, narrative, fast, smooth',
+      },
+      {
+        value: 'harrison',
+        name: 'Harrison - american, adult, male, neutral, narrative, fast, round',
+      },
+      {
+        value: 'hayden',
+        name: 'Cooper - american, adult, male, neutral, narrative, neutral, round',
+      },
+      {
+        value: 'hipolito',
+        name: 'Audrey - american, adult, female, low, narrative, slow, round',
+      },
+      {
+        value:
+          's3://mockingbird-prod/hook_1_chico_a3e5e83f-08ae-4a9f-825c-7e48d32d2fd8/voices/speaker/manifest.json',
+        name: 'Hook - american, male, gaming',
+      },
+      {
+        value: 's3://peregrine-voices/hudson saad parrot/manifest.json',
+        name: 'Hudson - american, adult, male, neutral, videos, neutral, thick',
+      },
+      {
+        value: 'hunter',
+        name: 'Hunter - british, old, male, high, narrative, fast, round',
+      },
+      {
+        value: 'ignacio',
+        name: 'Delilah - american, adult, female, neutral, narrative, slow, smooth',
+      },
+      {
+        value: 's3://peregrine-voices/mel28/manifest.json',
+        name: 'Jack - american, adult, male,',
+      },
+      {
+        value: 'jarrett',
+        name: 'Jarrett - american, adult, male, low, advertising, slow, smooth',
+      },
+      {
+        value:
+          's3://voice-cloning-zero-shot/801a663f-efd0-4254-98d0-5c175514c3e8/jennifer/manifest.json',
+        name: 'Jennifer - american, adult, female,',
+      },
+      {
+        value: 'jerrell',
+        name: 'Jerrell - american, adult, male, low, narrative, neutral, round',
+      },
+      {
+        value: 'jordan',
+        name: 'Jordan - american, adult, male, neutral, training, slow, round',
+      },
+      {
+        value:
+          's3://voice-cloning-zero-shot/dc23bb38-f568-4323-b6fb-7d64f685b97a/joseph/manifest.json',
+        name: 'Joseph - american, adult, male,',
+      },
+      {
+        value: 'judson',
+        name: 'Judson - american, adult, male, low, narrative, slow, smooth',
+      },
+      {
+        value: 'lance',
+        name: 'Lance - british, adult, male, low, videos, neutral, smooth',
+      },
+      {
+        value: 'larry',
+        name: 'Larry - american, adult, male, neutral, narrative, neutral, smooth',
+      },
+      {
+        value: 's3://peregrine-voices/larry_ads3_parrot_saad/manifest.json',
+        name: 'Larry (Advertising) - american, adult, male, neutral, advertising, neutral, smooth',
+      },
+      {
+        value:
+          's3://mockingbird-prod/larry_vo_narrative_4bd5c1bd-f662-4a38-b5b9-76563f7b92ec/voices/speaker/manifest.json',
+        name: 'Larry (Narrative) - american, adult, male, neutral, narrative, neutral, smooth',
+      },
+      {
+        value: 'lillian',
+        name: 'Lillian - british, old, female, neutral, training, slow, round',
+      },
+      {
+        value: 'ahmed',
+        name: 'Logan - british, old, male, neutral, narrative, neutral, thick',
+      },
+      {
+        value: 'lottie',
+        name: 'Lottie - british, adult, female, low, narrative, slow, smooth',
+      },
+      {
+        value: 'lucius',
+        name: 'Lucius - british, adult, male, low, narrative, slow, smooth',
+      },
+      {
+        value: 'mickey',
+        name: 'Madelyn - british, adult, female, neutral, videos, fast, thick',
+      },
+      {
+        value:
+          's3://voice-cloning-zero-shot/09b5c0cc-a8f4-4450-aaab-3657b9965d0b/podcaster/manifest.json',
+        name: 'Matt - american, adult, male,',
+      },
+      {
+        value: 's3://peregrine-voices/mel21/manifest.json',
+        name: 'Melissa - american, adult, female,',
+      },
+      {
+        value: 'micah',
+        name: 'Micah - british, adult, female, neutral, narrative, neutral, smooth',
+      },
+      {
+        value:
+          's3://voice-cloning-zero-shot/7c339a9d-370f-4643-adf5-4134e3ec9886/mlae02/manifest.json',
+        name: 'Michael - american, adult, male,',
+      },
+      {
+        value: 'mickey',
+        name: 'Madelyn - british, adult, female, neutral, videos, fast, thick',
+      },
+      {
+        value:
+          's3://voice-cloning-zero-shot/7c38b588-14e8-42b9-bacd-e03d1d673c3c/nicole/manifest.json',
+        name: 'Nicole - american, adult, female,',
+      },
+      {
+        value: 's3://peregrine-voices/nolan saad parrot/manifest.json',
+        name: 'Nolan - british, adult, male, high, videos, neutral, round',
+      },
+      {
+        value: 'nova',
+        name: 'Nova - american, adult, female, whisper, narrative, slow, smooth',
+      },
+      {
+        value: 'oliver',
+        name: 'Oliver - british, adult, male, high, videos, neutral, round',
+      },
+      {
+        value: 'oscar',
+        name: 'Oscar - british, adult, male, neutral, narrative, slow, smooth',
+      },
+      {
+        value: 'owen',
+        name: 'Owen - american, youth, male, high, narrative, neutral, round',
+      },
+      {
+        value: 'pedro',
+        name: 'Pedro - american, adult, male, neutral, narrative, slow, round',
+      },
+      {
+        value: 'phoebe',
+        name: 'Phoebe - british, adult, female, high, videos, fast, smooth',
+      },
+      {
+        value: 'randall',
+        name: 'Randall - british, adult, male, high, narrative, fast, thick',
+      },
+      {
+        value: 'reynaldo',
+        name: 'Reynaldo - british, old, male, low, narrative, fast, smooth',
+      },
+      {
+        value: 'rodrick',
+        name: 'Rodrick - american, adult, male, neutral, narrative, neutral, smooth',
+      },
+      {
+        value: 'gabriel',
+        name: 'Samantha - american, old, female, neutral, narrative, neutral, thick',
+      },
+      {
+        value: 'samuel',
+        name: 'Samuel - american, old, male, high, narrative, slow, gravelly',
+      },
+      {
+        value:
+          // eslint-disable-next-line max-len
+          's3://mockingbird-prod/agent_47_carmelo_pampillonio_58e796e1-0b87-4f3e-8b36-7def6d65ce66/voices/speaker/manifest.json',
+        name: 'Sarge - american, male, gaming',
+      },
+      {
+        value:
+          's3://voice-cloning-zero-shot/1f44b3e7-22ea-4c2e-87d0-b4d9c8f1d47d/sophia/manifest.json',
+        name: 'Sophia - american, adult, female,',
+      },
+      {
+        value: 'spencer',
+        name: 'April - british, adult, female, neutral, narrative, slow, smooth',
+      },
+      {
+        value: 'stella',
+        name: 'Stella - british, old, female, neutral, training, slow, round',
+      },
+      {
+        value: 'susan',
+        name: 'Susan - american, adult, female, high, videos, neutral, round',
+      },
+      {
+        value:
+          // eslint-disable-next-line max-len
+          's3://mockingbird-prod/susan_vo_commercials_0f4fa663-6eba-4582-be1e-2d5bde798f1c/voices/speaker/manifest.json',
+        name: 'Susan (Advertising) - american, adult, female, high, advertising, neutral, round',
+      },
+      {
+        value:
+          's3://mockingbird-prod/susan_vo_narrative_73051c90-460b-4e54-adab-9235f45c5e5f/voices/speaker/manifest.json',
+        name: 'Susan (Narrative) - american, adult, female, high, narrative, neutral, round',
+      },
+      {
+        value:
+          's3://mockingbird-prod/susan_vo_training_46ffcc60-d630-42f6-acfe-4affd003ae7a/voices/speaker/manifest.json',
+        name: 'Susan (Training) - american, adult, female, high, training, neutral, round',
+      },
+      {
+        value: 'theodore',
+        name: 'Theodore - american, old, male, neutral, narrative, neutral, gravelly',
+      },
+      {
+        value: 'victor',
+        name: 'Ariana - american, youth, female, high, videos, fast, thick',
+      },
+      {
+        value: 'wilbert',
+        name: 'Wilbert - british, adult, male, neutral, narrative, neutral, round',
+      },
+      {
+        value: 'wilbur',
+        name: 'Wilbur - american, youth, male, neutral, narrative, neutral, smooth',
+      },
+      {
+        value: 'wilfred',
+        name: 'Wilfred - american, old, male, low, training, slow, smooth',
+      },
+      {
+        value: 's3://peregrine-voices/mel22/manifest.json',
+        name: 'Will - american, adult, male,',
+      },
+      {
+        value: 'william',
+        name: 'William - american, adult, male, neutral, videos, neutral, round',
+      },
+      {
+        value:
+          // eslint-disable-next-line max-len
+          's3://mockingbird-prod/william_vo_narrative_0eacdff5-6243-4e26-8b3b-66e03458c1d1/voices/speaker/manifest.json',
+        name: 'William (Narrative) - american, adult, male, neutral, narrative, neutral, round',
+      },
+      {
+        value:
+          's3://mockingbird-prod/william_vo_training_1b939b71-14fa-41f0-b1db-7d94f194ad0a/voices/speaker/manifest.json',
+        name: 'William (Training) - american, adult, male, neutral, training, neutral, round',
+      },
+    ],
+  },
+  {
+    value: 'en-GB',
+    name: 'English (GB)',
+    voices: [
+      {
+        value: 's3://peregrine-voices/arthur ads parrot saad/manifest.json',
+        name: 'Arthur (Advertising) - british, adult, male, neutral, advertising, neutral, smooth',
+      },
+      {
+        value:
+          // eslint-disable-next-line max-len
+          's3://mockingbird-prod/arthur_vo_meditatoin_211f702d-b185-4115-b8b4-801f8130a38d/voices/speaker/manifest.json',
+        name: 'Arthur (Meditation) - british, adult, male, neutral, meditation, neutral, smooth',
+      },
+      {
+        value:
+          's3://mockingbird-prod/arthur_vo_narrative_a33fd610-73a9-4401-9a78-6b8219c68a9e/voices/speaker/manifest.json',
+        name: 'Arthur (Narrative) - british, adult, male, neutral, narrative, neutral, smooth',
+      },
+      {
+        value:
+          's3://mockingbird-prod/arthur_vo_training_9281c8fd-c7f0-4445-a148-466292d3d329/voices/speaker/manifest.json',
+        name: 'Arthur (Training) - british, adult, male, neutral, training, neutral, smooth',
+      },
+      {
+        value:
+          's3://mockingbird-prod/eileen_vo_5d7b2bcc-d635-4301-97e8-d97c13768514/voices/speaker/manifest.json',
+        name: 'Eileen - british, female, narrative',
+      },
+      {
+        value: 'frankie',
+        name: 'Frankie - british, old, male, neutral, training, neutral, thick',
+      },
+      {
+        value:
+          's3://voice-cloning-zero-shot/418a94fa-2395-4487-81d8-22daf107781f/george/manifest.json',
+        name: 'George - british, adult, male,',
+      },
+      {
+        value: 'julian',
+        name: 'Julian - british, adult, male, neutral, videos, neutral, round',
+      },
+      {
+        value:
+          's3://voice-cloning-zero-shot/0b5b2e4b-5103-425e-8aa0-510dd35226e2/mark/manifest.json',
+        name: 'Mark - british, adult, male,',
+      },
+      {
+        value: 's3://peregrine-voices/oliver_ads2_parrot_saad/manifest.json',
+        name: 'Oliver (Advertising) - british, adult, male, high, advertising, neutral, round',
+      },
+      {
+        value:
+          's3://peregrine-voices/oliver_narrative2_parrot_saad/manifest.json',
+        name: 'Oliver (Narrative) - british, adult, male, high, narrative, neutral, round',
+      },
+      {
+        value:
+          's3://mockingbird-prod/oliver_vo_training_6e3f604a-5605-4542-948d-347b0d7546fc/voices/speaker/manifest.json',
+        name: 'Oliver (Training) - british, adult, male, high, training, neutral, round',
+      },
+      {
+        value:
+          's3://voice-cloning-zero-shot/820da3d2-3a3b-42e7-844d-e68db835a206/sarah/manifest.json',
+        name: 'Sarah - british, adult, female,',
+      },
+    ],
+  },
+  {
+    value: 'en-AU',
+    name: 'English (AU)',
+    voices: [
+      {
+        value: 's3://peregrine-voices/barry ads parrot saad/manifest.json',
+        name: 'Barry (Advertising) - australian, male, advertising',
+      },
+      {
+        value:
+          's3://peregrine-voices/barry narrative parrot saad/manifest.json',
+        name: 'Barry (Narrative) - australian, male, narrative',
+      },
+      {
+        value: 'frederick',
+        name: 'Frederick - australian, adult, male, low, narrative, slow, thick',
+      },
+      {
+        value: 's3://peregrine-voices/russell2_parrot_saad/manifest.json',
+        name: 'Russell - australian, male,',
+      },
+    ],
+  },
+  {
+    value: 'en-CA',
+    name: 'English (CA)',
+    voices: [
+      {
+        value: 's3://peregrine-voices/charlotte ads parrot saad/manifest.json',
+        name: 'Charlotte (Advertising) - canadian, adult, female, low, advertising, neutral, smooth',
+      },
+      {
+        value:
+          's3://peregrine-voices/charlotte meditation 2 parrot saad/manifest.json',
+        name: 'Charlotte (Meditation) - canadian, adult, female, low, meditation, neutral, smooth',
+      },
+      {
+        value:
+          // eslint-disable-next-line max-len
+          's3://mockingbird-prod/charlotte_vo_narrative_9290be17-ccea-4700-a7fd-a8fe5c49fb20/voices/speaker/manifest.json',
+        name: 'Charlotte (Narrative) - canadian, adult, female, low, narrative, neutral, smooth',
+      },
+      {
+        value:
+          's3://peregrine-voices/charlotte_training_parrot_saad/manifest.json',
+        name: 'Charlotte (Training) - canadian, adult, female, low, training, neutral, smooth',
+      },
+      {
+        value:
+          // eslint-disable-next-line max-len
+          's3://mockingbird-prod/olivia_vo_commercials_6e3c384f-15d6-4fe7-b9a4-0cb1d69daeba/voices/speaker/manifest.json',
+        name: 'Olivia (Advertising) - canadian, female, advertising',
+      },
+      {
+        value: 's3://peregrine-voices/olivia_ads3_parrot_saad/manifest.json',
+        name: 'Olivia (Narrative) - canadian, female, narrative',
+      },
+      {
+        value:
+          's3://mockingbird-prod/olivia_vo_training_4376204f-a411-4e5d-a5c0-ce6cc3908052/voices/speaker/manifest.json',
+        name: 'Olivia (Training) - canadian, female, training',
+      },
+    ],
+  },
+  {
+    value: 'en-IE',
+    name: 'English (IE)',
+    voices: [
+      {
+        value: 'florencio',
+        name: 'Madison - irish, old, female, neutral, narrative, slow, round',
+      },
+    ],
+  },
+  {
+    value: 'en-NZ',
+    name: 'English (NZ)',
+    voices: [
+      {
+        value:
+          's3://voice-cloning-zero-shot/d9ff78ba-d016-47f6-b0ef-dd630f59414e/female-cs/manifest.json',
+        name: 'Ruby - australian, adult, female,',
+      },
+    ],
+  },
+];

lib/utils/speech-utils.js

@@ -17,10 +17,13 @@ const TtsIbmLanguagesVoices = require('./speech-data/tts-ibm');
 const TtsNvidiaLanguagesVoices = require('./speech-data/tts-nvidia');
 const TtsElevenlabsLanguagesVoices = require('./speech-data/tts-elevenlabs');
 const TtsWhisperLanguagesVoices = require('./speech-data/tts-whisper');
+const TtsPlayHtLanguagesVoices = require('./speech-data/tts-playht');
 
 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 TtsModelRimelabs = require('./speech-data/tts-model-rimelabs');
 
 const SttGoogleLanguagesVoices = require('./speech-data/stt-google');
 const SttAwsLanguagesVoices = require('./speech-data/stt-aws');
@@ -162,16 +165,26 @@ const testAwsTts = async(logger, getTtsVoices, credentials) => {
   }
 };
 
-const testAwsStt = async(logger, credentials) => {
+const testAwsStt = async(logger, getAwsAuthToken, credentials) => {
   try {
-    const {region, accessKeyId, secretAccessKey} = credentials;
-    const client = new TranscribeClient({
-      region,
-      credentials: {
-        accessKeyId,
-        secretAccessKey
-      }
-    });
+    const {region, accessKeyId, secretAccessKey, roleArn} = credentials;
+    let client = null;
+    if (accessKeyId && secretAccessKey) {
+      client = new TranscribeClient({
+        region,
+        credentials: {
+          accessKeyId,
+          secretAccessKey
+        }
+      });
+    } else if (roleArn) {
+      client = new TranscribeClient({
+        region,
+        credentials: await getAwsAuthToken(null, null, region, roleArn),
+      });
+    } else {
+      client = new TranscribeClient({region});
+    }
     const command = new ListVocabulariesCommand({});
     const response =  await client.send(command);
     return response;
@@ -240,6 +253,50 @@ 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!'
+      }
+    );
+    // Test if playHT can fetch voices
+    await fetchLayHTVoices(credentials);
+  } catch (err) {
+    logger.info({err}, 'synth Playht returned error');
+    throw err;
+  }
+};
+
+const testRimelabs = async(logger, synthAudio, credentials) => {
+  try {
+    await synthAudio(
+      {
+        increment: () => {},
+        histogram: () => {}
+      },
+      {
+        vendor: 'rimelabs',
+        credentials,
+        language: 'en-US',
+        voice: 'amber',
+        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: () => {}},
@@ -263,9 +320,7 @@ const testDeepgramTTS = async(logger, synthAudio, credentials) => {
       {
         vendor: 'deepgram',
         credentials,
-        language: 'en-US',
-        voice: 'alpha-aurora-en-v2',
-        model: 'alpha-aurora-en-v2',
+        model: 'aura-asteria-en',
         text: 'Hi there and welcome to jambones!'
       }
     );
@@ -355,6 +410,7 @@ const getSpeechCredential = (credential, logger) => {
       ...credential,
       accessKeyId: credential.access_key_id,
       secretAccessKey: credential.secret_access_key,
+      roleArn: credential.role_arn,
       region: credential.aws_region || 'us-east-1'
     };
   }
@@ -376,6 +432,7 @@ function decryptCredential(obj, credential, logger, isObscureKey = true) {
   else if ('aws' === obj.vendor) {
     const o = JSON.parse(decrypt(credential));
     obj.access_key_id = o.access_key_id;
+    obj.role_arn = o.role_arn;
     obj.secret_access_key = isObscureKey ? obscureKey(o.secret_access_key) : o.secret_access_key;
     obj.aws_region = o.aws_region;
     logger.info({obj, o}, 'retrieving aws speech credential');
@@ -430,7 +487,17 @@ 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;
-    obj.use_streaming = o.use_streaming;
+  } 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 ('rimelabs' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+    obj.model_id = o.model_id;
+    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;
@@ -491,6 +558,10 @@ 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 'rimelabs':
+      return await getLanguagesVoicesForRimelabs(credential, getTtsVoices, logger);
     case 'assemblyai':
       return await getLanguagesVoicesForAssemblyAI(credential, getTtsVoices, logger);
     case 'whisper':
@@ -525,6 +596,7 @@ async function getLanguagesVoicesForAws(credential, getTtsVoices, logger) {
         credentials: {
           accessKeyId: credential.access_key_id,
           secretAccessKey: credential.secret_access_key,
+          roleArn: credential.role_arn,
           region: credential.aws_region || process.env.AWS_REGION
         }
       });
@@ -648,6 +720,92 @@ async function getLanguagesVoicesForElevenlabs(credential) {
   }
 }
 
+const concat = (a) => {
+  return a ? ` ${a},` : '';
+};
+
+const fetchLayHTVoices = async(credential) => {
+  if (credential) {
+    const get = bent('https://api.play.ht', 'GET', 'json', {
+      'AUTHORIZATION' : credential.api_key,
+      'X-USER-ID': credential.user_id,
+      'Accept': 'application/json'
+    });
+
+    const voices = await get('/api/v2/voices');
+    let clone_voices = [];
+    try {
+      // try if the account has permission to cloned voice
+      //otherwise ignore this.
+      clone_voices = await get('/api/v2/cloned-voices');
+    } catch {}
+    return [clone_voices, voices];
+  }
+};
+
+async function getLanguagesVoicesForPlayHT(credential) {
+  if (credential) {
+    const [cloned_voice, voices] = await fetchLayHTVoices(credential);
+    const list_voices = [...cloned_voice, ...voices];
+
+    const buildVoice = (d) => {
+      let name = `${d.name} -${concat(d.accent)}${concat(d.age)}${concat(d.gender)}${concat(d.loudness)}` +
+      `${concat(d.style)}${concat(d.tempo)}${concat(d.texture)}` ;
+      name = name.endsWith(',') ? name.trim().slice(0, -1) : name;
+      return {
+        value: `${d.id}`,
+        name
+      };
+    };
+
+    const ttsVoices = list_voices.reduce((acc, voice) => {
+      if (!credential.voice_engine.includes(voice.voice_engine)) {
+        return acc;
+      }
+      const languageCode = voice.language_code;
+      // custom voice does not have language code
+      if (!languageCode) {
+        voice.language_code = 'en';
+        voice.language = 'Custom-English';
+      }
+      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(TtsPlayHtLanguagesVoices, undefined, TtsModelPlayHT);
+}
+
+async function getLanguagesVoicesForRimelabs(credential) {
+  const model_id = credential ? credential.model_id : null;
+  const get = bent('https://users.rime.ai', 'GET', 'json', {
+    'Accept': 'application/json'
+  });
+  const voices = await get('/data/voices/all.json');
+  let selectedVoices = model_id ? voices[model_id] : Object.values(voices).reduce((acc, val) => [...acc, ...val], []);
+  selectedVoices = selectedVoices.map((v) => ({
+    name: v.charAt(0).toUpperCase() + v.slice(1),
+    value: v
+  }));
+  const ttsVoices = [
+    {
+      value: 'en-US',
+      name: 'English (US)',
+      voices: selectedVoices
+    }
+  ];
+  return tranform(ttsVoices, undefined, TtsModelRimelabs);
+}
+
 async function getLanguagesVoicesForAssemblyAI(credential) {
   return tranform(undefined, SttAssemblyaiLanguagesVoices);
 }
@@ -799,6 +957,8 @@ module.exports = {
   testIbmStt,
   testSonioxStt,
   testElevenlabs,
+  testPlayHT,
+  testRimelabs,
   testAssemblyStt,
   testDeepgramTTS,
   getSpeechCredential,

package.json

@@ -1,6 +1,6 @@
 {
   "name": "jambonz-api-server",
-  "version": "0.8.5",
+  "version": "0.9.0",
   "description": "",
   "main": "app.js",
   "scripts": {
@@ -19,54 +19,54 @@
     "url": "https://github.com/jambonz/jambonz-api-server.git"
   },
   "dependencies": {
-    "@aws-sdk/client-s3": "^3.363.0",
-    "@aws-sdk/client-transcribe": "^3.363.0",
-    "@azure/storage-blob": "^12.15.0",
+    "@aws-sdk/client-s3": "^3.550.0",
+    "@aws-sdk/client-transcribe": "^3.549.0",
+    "@azure/storage-blob": "^12.17.0",
     "@deepgram/sdk": "^1.21.0",
-    "@google-cloud/speech": "^5.2.0",
-    "@google-cloud/storage": "^6.12.0",
+    "@google-cloud/speech": "^6.5.0",
+    "@google-cloud/storage": "^7.9.0",
     "@jambonz/db-helpers": "^0.9.3",
     "@jambonz/lamejs": "^1.2.2",
     "@jambonz/mw-registrar": "^0.2.7",
-    "@jambonz/realtimedb-helpers": "^0.8.7",
-    "@jambonz/speech-utils": "^0.0.41",
+    "@jambonz/realtimedb-helpers": "^0.8.8",
+    "@jambonz/speech-utils": "^0.1.0",
     "@jambonz/time-series": "^0.2.8",
-    "@jambonz/verb-specifications": "^0.0.45",
-    "@soniox/soniox-node": "^1.1.1",
-    "argon2": "^0.30.3",
-    "assemblyai": "^3.0.1",
+    "@jambonz/verb-specifications": "^0.0.69",
+    "@soniox/soniox-node": "^1.2.2",
+    "argon2": "^0.40.1",
+    "assemblyai": "^4.3.4",
     "bent": "^7.3.12",
     "cors": "^2.8.5",
     "debug": "^4.3.4",
-    "express": "^4.18.1",
-    "express-rate-limit": "^6.4.0",
-    "form-data": "^2.5.1",
-    "helmet": "^5.1.0",
-    "ibm-watson": "^7.1.2",
-    "jsonwebtoken": "^9.0.0",
-    "mailgun.js": "^9.1.2",
-    "microsoft-cognitiveservices-speech-sdk": "1.31.0",
-    "mysql2": "^2.3.3",
-    "nocache": "3.0.4",
-    "passport": "^0.6.0",
+    "express": "^4.19.2",
+    "express-rate-limit": "^7.2.0",
+    "form-data": "^4.0.0",
+    "helmet": "^7.1.0",
+    "ibm-watson": "^9.0.1",
+    "jsonwebtoken": "^9.0.2",
+    "mailgun.js": "^10.2.1",
+    "microsoft-cognitiveservices-speech-sdk": "1.36.0",
+    "mysql2": "^3.9.3",
+    "nocache": "4.0.0",
+    "passport": "^0.7.0",
     "passport-http-bearer": "^1.0.1",
-    "pino": "^5.17.0",
-    "short-uuid": "^4.1.0",
+    "pino": "^8.20.0",
+    "short-uuid": "^4.2.2",
     "stream-buffers": "^3.0.2",
-    "stripe": "^8.222.0",
-    "swagger-ui-express": "^4.4.0",
-    "uuid": "^8.3.2",
+    "stripe": "^14.24.0",
+    "swagger-ui-express": "^5.0.0",
+    "uuid": "^9.0.1",
     "wav": "^1.0.2",
-    "ws": "^8.12.1",
+    "ws": "^8.16.0",
     "yamljs": "^0.3.0"
   },
   "devDependencies": {
     "eslint": "^8.39.0",
     "eslint-plugin-promise": "^6.1.1",
-    "husky": "7.0.4",
+    "husky": "9.0.11",
     "nyc": "^15.1.0",
     "request": "^2.88.2",
     "request-promise-native": "^1.0.9",
-    "tape": "^5.5.3"
+    "tape": "^5.7.5"
   }
 }

test/docker-compose-testbed.yaml

@@ -10,7 +10,7 @@ networks:
 
 services:  
   mysql:
-    platform: linux/x86_64
+    # platform: linux/x86_64
     image: mysql:5.7
     ports:
       - "3360:3306"
@@ -36,7 +36,7 @@ services:
         ipv4_address: 172.58.0.3
     
   influxdb:
-    platform: linux/x86_64
+    # platform: linux/x86_64
     image: influxdb:1.8
     ports:
       - "8086:8086"

test/speech-credentials.js

@@ -170,6 +170,20 @@ test('speech credentials tests', async(t) => {
       //console.log(JSON.stringify(result));
       t.ok(result.statusCode === 200 && result.body.tts.status === 'ok', 'successfully tested speech credential for google tts');
       t.ok(result.statusCode === 200 && result.body.stt.status === 'ok', 'successfully tested speech credential for google stt');
+
+      result = await request.post(`/Accounts/${account_sid}/TtsCache/Synthesize`, {
+        resolveWithFullResponse: true,
+        auth: authUser,
+        json: true,
+        body: {
+          speech_credential_sid: ms_sid,
+          text: "Hello How are you",
+          language: "en-US",
+          voice: "en-US-Standard-C"
+        }
+      });
+
+      t.ok(result.statusCode === 200, 'successfully google tested synthesize');
     }
 
     /* add / test a credential for microsoft */
@@ -198,6 +212,20 @@ test('speech credentials tests', async(t) => {
       //console.log(JSON.stringify(result));
       t.ok(result.statusCode === 200 && result.body.tts.status === 'ok', 'successfully tested speech credential for microsoft tts');
       t.ok(result.statusCode === 200 && result.body.stt.status === 'ok', 'successfully tested speech credential for microsoft stt');
+
+      result = await request.post(`/Accounts/${account_sid}/TtsCache/Synthesize`, {
+        resolveWithFullResponse: true,
+        auth: authUser,
+        json: true,
+        body: {
+          speech_credential_sid: ms_sid,
+          text: "Hello How are you",
+          language: "en-US",
+          voice: "en-US-AvaMultilingualNeural"
+        }
+      });
+
+      t.ok(result.statusCode === 200, 'successfully microsoft tested synthesize');
     }
 
     /* add / test a credential for AWS */
@@ -227,6 +255,20 @@ test('speech credentials tests', async(t) => {
       //console.log(JSON.stringify(result));
       t.ok(result.statusCode === 200 && result.body.tts.status === 'ok', 'successfully tested speech credential for AWS tts');
       t.ok(result.statusCode === 200 && result.body.stt.status === 'ok', 'successfully tested speech credential for AWS stt');
+
+      result = await request.post(`/Accounts/${account_sid}/TtsCache/Synthesize`, {
+        resolveWithFullResponse: true,
+        auth: authUser,
+        json: true,
+        body: {
+          speech_credential_sid: ms_sid,
+          text: "Hello How are you",
+          language: "en-US",
+          voice: "Joanna"
+        }
+      });
+
+      t.ok(result.statusCode === 200, 'successfully AWS tested synthesize');
     }
 
     /* add a credential for wellsaid */
@@ -253,6 +295,20 @@ test('speech credentials tests', async(t) => {
       //console.log(JSON.stringify(result));
       t.ok(result.statusCode === 200 && result.body.tts.status === 'ok', 'successfully tested speech credential for wellsaid');
 
+      result = await request.post(`/Accounts/${account_sid}/TtsCache/Synthesize`, {
+        resolveWithFullResponse: true,
+        auth: authUser,
+        json: true,
+        body: {
+          speech_credential_sid: ms_sid,
+          text: "Hello How are you",
+          language: "en-US",
+          voice: "3"
+        }
+      });
+
+      t.ok(result.statusCode === 200, 'successfully Wellsaid tested synthesize');
+
       /* delete the credential */
       result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${ms_sid}`, {
         auth: authUser,
@@ -285,6 +341,20 @@ test('speech credentials tests', async(t) => {
       //console.log(JSON.stringify(result));
       t.ok(result.statusCode === 200 && result.body.stt.status === 'ok', 'successfully tested speech credential for deepgram');
 
+      result = await request.post(`/Accounts/${account_sid}/TtsCache/Synthesize`, {
+        resolveWithFullResponse: true,
+        auth: authUser,
+        json: true,
+        body: {
+          speech_credential_sid: ms_sid,
+          text: "Hello How are you",
+          language: "en-US",
+          voice: "aura-asteria-en"
+        }
+      });
+
+      t.ok(result.statusCode === 200, 'successfully deepgram tested synthesize');
+
       /* delete the credential */
       result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${ms_sid}`, {
         auth: authUser,
@@ -404,6 +474,20 @@ test('speech credentials tests', async(t) => {
       //console.log(JSON.stringify(result));
       t.ok(result.statusCode === 200 && result.body.stt.status === 'ok', 'successfully tested speech credential for ibm stt');
 
+      result = await request.post(`/Accounts/${account_sid}/TtsCache/Synthesize`, {
+        resolveWithFullResponse: true,
+        auth: authUser,
+        json: true,
+        body: {
+          speech_credential_sid: ms_sid,
+          text: "Hello How are you",
+          language: "en-US",
+          voice: "en-US_MichaelExpressive"
+        }
+      });
+
+      t.ok(result.statusCode === 200, 'successfully IBM tested synthesize');
+
       /* delete the credential */
       result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${ms_sid}`, {
         auth: authUser,
@@ -536,7 +620,7 @@ test('speech credentials tests', async(t) => {
         model_id: 'eleven_multilingual_v2'
       }
     });
-    t.ok(result.statusCode === 201, 'successfully added speech credential for Cobalt');
+    t.ok(result.statusCode === 201, 'successfully added speech credential for elevenlabs');
     const elevenlabs_sid = result.body.sid;
 
     /* delete the credential */
@@ -544,7 +628,54 @@ test('speech credentials tests', async(t) => {
       auth: authUser,
       resolveWithFullResponse: true,
     });
-    t.ok(result.statusCode === 204, 'successfully deleted speech credential for Cobalt');
+    t.ok(result.statusCode === 204, 'successfully deleted speech credential for elevenlabs');
+
+    /* add a credential for playht */
+    result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
+      resolveWithFullResponse: true,
+      auth: authUser,
+      json: true,
+      body: {
+        vendor: 'playht',
+        use_for_stt: false,
+        use_for_tts: true,
+        api_key: 'asdasdasdasddsadasda',
+        user_id: 'user_id',
+        voice_engine: 'PlayHT2.0-turbo'
+      }
+    });
+    t.ok(result.statusCode === 201, 'successfully added speech credential for playht');
+    const playht_sid = result.body.sid;
+
+    /* delete the credential */
+    result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${playht_sid}`, {
+      auth: authUser,
+      resolveWithFullResponse: true,
+    });
+    t.ok(result.statusCode === 204, 'successfully deleted speech credential for playht');
+
+    /* add a credential for rimelabs */
+    result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
+      resolveWithFullResponse: true,
+      auth: authUser,
+      json: true,
+      body: {
+        vendor: 'rimelabs',
+        use_for_stt: false,
+        use_for_tts: true,
+        api_key: 'asdasdasdasddsadasda',
+        model_id: 'mist',
+      }
+    });
+    t.ok(result.statusCode === 201, 'successfully added speech credential for rimelabs');
+    const rimelabs_sid = result.body.sid;
+
+    /* delete the credential */
+    result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${rimelabs_sid}`, {
+      auth: authUser,
+      resolveWithFullResponse: true,
+    });
+    t.ok(result.statusCode === 204, 'successfully deleted speech credential for rimelabs');
 
 
     /* add a credential for custom voices google */
@@ -612,6 +743,30 @@ test('speech credentials tests', async(t) => {
     });
     t.ok(result.statusCode === 204, 'successfully deleted speech credential');
 
+    /* add a credential for aws polly by roleArn */
+    result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
+      resolveWithFullResponse: true,
+      auth: authUser,
+      json: true,
+      body: {
+        vendor: 'aws',
+        labe: 'aws_polly_with_arn',
+        use_for_tts: true,
+        use_for_stt: false,
+        role_arn: 'Arn::aws::role',
+        aws_region: 'us-east-1'
+      }
+    });
+    t.ok(result.statusCode === 201, 'successfully added speech credential for AWS Polly By RoleArn');
+    const awsPollySid = result.body.sid;
+
+    /* delete the credential */
+    result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${awsPollySid}`, {
+      auth: authUser,
+      resolveWithFullResponse: true,
+    });
+    t.ok(result.statusCode === 204, 'successfully deleted speech credential');
+
     /* Check google supportedLanguagesAndVoices */
     result = await request.get(`/Accounts/${account_sid}/SpeechCredentials/speech/supportedLanguagesAndVoices?vendor=google`, {
       resolveWithFullResponse: true,