ben.carrasco/speech-utils
1
0
Fork 0
mirror of https://github.com/jambonz/speech-utils.git synced 2026-01-25 02:08:26 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #113 from vasudevanubrolu/feat/893-azure-ssml

Browse Source 
Feat/893 azure ssml
This commit is contained in:
Dave Horton
2025-05-27 10:08:48 -04:00
committed by GitHub
parent 0d98f73c43 925bd26a70
commit 2ff8587601
2 changed files with 18 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
lib/config.js
View File

@@ -2,6 +2,7 @@ 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;
@@ -21,5 +22,6 @@ module.exports = {
JAMBONES_TTS_CACHE_DURATION_MINS,
JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
TMP_FOLDER,
HTTP_TIMEOUT
HTTP_TIMEOUT,
JAMBONES_AZURE_ENABLE_SSML
};

22
lib/synth-audio.js
View File

@@ -47,6 +47,7 @@ const {
JAMBONES_HTTP_PROXY_PORT,
JAMBONES_TTS_CACHE_DURATION_MINS,
JAMBONES_TTS_TRIM_SILENCE,
JAMBONES_AZURE_ENABLE_SSML
} = require('./config');
const EXPIRES = JAMBONES_TTS_CACHE_DURATION_MINS;
const OpenAI = require('openai');
@@ -466,7 +467,6 @@ async function _synthOnPremMicrosoft(logger, {
}) {
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
@@ -479,9 +479,13 @@ async function _synthOnPremMicrosoft(logger, {
/* 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}">${words}</voice></speak>`;
content = `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="${language}"><lang xml:lang="${language}"><voice name="${voice}">${words}</voice></lang></speak>`;
logger.info({content}, 'synthMicrosoft');
}
else if (JAMBONES_AZURE_ENABLE_SSML) {
// eslint-disable-next-line max-len
content = `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="${language}"><voice name="${voice}"><lang xml:lang="${language}">${text}</lang></voice></speak>`;
}
try {
const trimSilence = JAMBONES_TTS_TRIM_SILENCE;
@@ -518,19 +522,23 @@ const synthMicrosoft = async(logger, {
let content = text;
if (use_custom_tts && !content.startsWith('<speak')) {
/**
* Note: it seems that to use custom voice ssml is required with the voice attribute
* Otherwise sending plain text we get "Voice does not match"
*/
* Note: it seems that to use custom voice ssml is required with the voice attribute
* Otherwise sending plain text we get "Voice does not match"
*/
content = `<speak>${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}">${words}</voice></speak>`;
content = `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="${language}"><lang xml:lang="${language}"><voice name="${voice}">${words}</voice></lang></speak>`;
logger.info({content}, 'synthMicrosoft');
}
else if (JAMBONES_AZURE_ENABLE_SSML) {
// eslint-disable-next-line max-len
content = `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="${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 = '';