Compare commits

...
8 Commits
Author SHA1 Message Date
Dave Horton 7dc3bbdb01 0.2.8 2025-05-08 09:31:40 -04:00
Dave Horton f8f9de2645 Merge pull request #110 from jambonz/feat/rimlabs_arcana
support rimelabs arcana
2025-05-06 09:22:35 -04:00
Quan HL 467d7ede26 support rimelabs arcana 2025-05-06 16:16:18 +07:00
Dave Horton fc211ab2e7 0.2.7 2025-04-28 19:31:59 -04:00
Dave Horton 536f8aab31 Merge pull request #109 from jambonz/feat/riva_tts
support riva tts stream
2025-04-28 19:31:30 -04:00
Hoan Luu Huu 69b3fdffbe Merge branch 'main' into feat/riva_tts 2025-04-28 08:47:44 +07:00
Hoan Luu Huu a4a427e174 Merge branch 'main' into feat/riva_tts 2025-04-23 18:11:19 +07:00
Quan HL d35d7f0aec support riva tts stream 2025-04-23 17:54:21 +07:00
3 changed files with 27 additions and 5 deletions
+24 -2
View File
@@ -215,7 +215,8 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
audioData = await synthNuance(client, logger, {credentials, stats, voice, model, text});
break;
case 'nvidia':
audioData = await synthNvidia(client, logger, {credentials, stats, language, voice, model, text});
audioData = await synthNvidia(client, logger, {credentials, stats, language, voice, model, text,
renderForCaching, disableTtsStreaming});
break;
case 'ibm':
audioData = await synthIbm(logger, {credentials, stats, voice, text});
@@ -708,8 +709,24 @@ const synthNuance = async(client, logger, {credentials, stats, voice, model, tex
});
};
const synthNvidia = async(client, logger, {credentials, stats, language, voice, model, text}) => {
const synthNvidia = async(client, logger, {
credentials, stats, language, voice, model, text, renderForCaching, disableTtsStreaming
}) => {
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}${text.replace(/\n/g, ' ')}`,
servedFromCache: false,
rtt: 0
};
}
let rivaClient, request;
const sampleRate = 8000;
try {
@@ -946,6 +963,11 @@ const synthRimelabs = async(logger, {
params += ',write_cache_file=1';
if (opts.speedAlpha) params += `,speed_alpha=${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 {
+2 -2
View File
@@ -1,12 +1,12 @@
{
"name": "@jambonz/speech-utils",
"version": "0.2.6",
"version": "0.2.8",
"lockfileVersion": 2,
"requires": true,
"packages": {
"": {
"name": "@jambonz/speech-utils",
"version": "0.2.6",
"version": "0.2.8",
"license": "MIT",
"dependencies": {
"@aws-sdk/client-polly": "^3.496.0",
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@jambonz/speech-utils",
"version": "0.2.6",
"version": "0.2.8",
"description": "TTS-related speech utilities for jambonz",
"main": "index.js",
"author": "Dave Horton",