From 5328a60de86c7887655bac498cf91a1138fa0da9 Mon Sep 17 00:00:00 2001
From: Quan HL <quan.luuhoang8@gmail.com>
Date: Wed, 3 Sep 2025 06:35:20 +0700
Subject: [PATCH] support custom tts Stream

---
 lib/synth-audio.js | 22 ++++++++++++++++++++--
 test/synth.js      |  3 +++
 2 files changed, 23 insertions(+), 2 deletions(-)

diff --git a/lib/synth-audio.js b/lib/synth-audio.js
index 7130972..2fb1cf6 100644
--- a/lib/synth-audio.js
+++ b/lib/synth-audio.js
@@ -271,7 +271,7 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
         break;
       case vendor.startsWith('custom') ? vendor : 'cant_match_value':
         audioData = await synthCustomVendor(logger,
-          {credentials, stats, language, voice, key, text});
+          {credentials, stats, language, voice, key, text, renderForCaching, disableTtsStreaming});
         break;
       default:
         assert(`synthAudio: unsupported speech vendor ${vendor}`);
@@ -818,9 +818,27 @@ const synthNvidia = async(client, logger, {
 };
 
 
-const synthCustomVendor = async(logger, {credentials, stats, language, voice, text, filePath}) => {
+const synthCustomVendor = async(logger, {credentials, stats, language, voice,
+  text, filePath, renderForCaching, disableTtsStreaming, key}) => {
   const {vendor, auth_token, custom_tts_url} = credentials;
 
+  if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
+    let params = '{';
+    params += `auth_token=${auth_token}`;
+    params += `,playback_id=${key}`;
+    params += `,custom_tts_url=${custom_tts_url}`;
+    params += ',vendor=custom';
+    params += `,voice=${voice}`;
+    params += ',write_cache_file=1';
+    params += '}';
+
+    return {
+      filePath: `say:${params}${text.replace(/\n/g, ' ').replace(/\r/g, ' ')}`,
+      servedFromCache: false,
+      rtt: 0
+    };
+  }
+
   try {
     const post = bent('POST', {
       'Authorization': `Bearer ${auth_token}`,
diff --git a/test/synth.js b/test/synth.js
index fddad81..8f088e4 100644
--- a/test/synth.js
+++ b/test/synth.js
@@ -554,6 +554,7 @@ test('Custom Vendor speech synth tests', async(t) => {
       language: 'en-US',
       voice: 'English-US.Female-1',
       text: 'This is a test.  This is only a test',
+      renderForCaching: true
     });
     t.ok(!opts.servedFromCache, `successfully synthesized custom vendor audio to ${opts.filePath}`);
     t.ok(opts.filePath.endsWith('wav'), 'audio is cached as wav file');
@@ -575,6 +576,7 @@ test('Custom Vendor speech synth tests', async(t) => {
       language: 'en-US',
       voice: 'English-US.Female-1',
       text: 'This is a test.  This is only a test',
+      renderForCaching: true
     });
     t.ok(opts.servedFromCache, `successfully get custom vendor cached audio to ${opts.filePath}`);
     t.ok(opts.filePath.endsWith('wav'), 'audio is cached as wav file');
@@ -589,6 +591,7 @@ test('Custom Vendor speech synth tests', async(t) => {
       language: 'en-US',
       voice: 'English-US.Female-1',
       text: '<speak>This is a test.  This is only a test</speak>',
+      renderForCaching: true
     });
     t.ok(!opts.servedFromCache, `successfully synthesized Custom Vendor audio to ${opts.filePath}`);
     obj = await getJSON(`http://127.0.0.1:3100/lastRequest/somethingnew2`);