ben.carrasco/jambonz-api-server
1
0
Fork 0
mirror of https://github.com/jambonz/jambonz-api-server.git synced 2026-07-24 04:52:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

support playht3.0 languages (#357)

Browse Source 
* support playht3.0 languages

* update speech utils version
This commit is contained in:
Hoan Luu Huu
2024-10-16 18:34:02 +07:00
committed by GitHub
parent 84d83a0a48
commit de250c8d58
5 changed files with 206 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/utils/speech-utils.js
+47 -25
View File
@@ -25,6 +25,7 @@ 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 ttsLanguagesPlayHt = require('./speech-data/tts-languages-playht');
const TtsModelRimelabs = require('./speech-data/tts-model-rimelabs');
const SttGoogleLanguagesVoices = require('./speech-data/stt-google');
@@ -40,6 +41,7 @@ const SttSpeechmaticsLanguagesVoices = require('./speech-data/stt-speechmatics')
const SttAssemblyaiLanguagesVoices = require('./speech-data/stt-assemblyai');
const SttVerbioLanguagesVoices = require('./speech-data/stt-verbio');
const testSonioxStt = async(logger, credentials) => {
const api_key = credentials;
const soniox = new SpeechClient(api_key);
@@ -869,6 +871,7 @@ const fetchLayHTVoices = async(credential) => {
async function getLanguagesVoicesForPlayHT(credential) {
if (credential) {
const {voice_engine} = credential;
const [cloned_voice, voices] = await fetchLayHTVoices(credential);
const list_voices = [...cloned_voice, ...voices];
@@ -876,38 +879,57 @@ async function getLanguagesVoicesForPlayHT(credential) {
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;
name += !d.language_code ? ' - Custom Voice' : '';
return {
value: `${d.id}`,
name
};
};
const ttsVoices = list_voices.reduce((acc, voice) => {
// Play3.0 support all voice for PlayHT2.0*
const filteredVoiceEngine = credential.voice_engine === 'Play3.0' ?
`${credential.voice_engine}_PlayHT2.0_PlayHT2.0-turbo` : credential.voice_engine;
if (!filteredVoiceEngine.includes(voice.voice_engine)) {
const buildPlay30Payload = () => {
// PlayHT3.0 can play different languages with differrent voice.
// all voices will be added to english language by default and orther langauges will get voices from english.
const ttsVoices = ttsLanguagesPlayHt.map((l) => ({
...l,
voices: l.value === 'english' ? list_voices.map((v) => buildVoice(v)) : []
}));
return tranform(ttsVoices, undefined, TtsModelPlayHT);
};
const buildPayload = () => {
const ttsVoices = list_voices.reduce((acc, voice) => {
if (!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;
}
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(ttsVoices, undefined, TtsModelPlayHT);
};
switch (voice_engine) {
case 'Play3.0':
return buildPlay30Payload();
default:
return buildPayload();
}
}
return tranform(TtsPlayHtLanguagesVoices, undefined, TtsModelPlayHT);
}