@jambonz/speech-utils 0.0.14 → 0.0.16

package/lib/synth-audio.js

@@ -30,9 +30,27 @@ const {
 const {SynthesizeSpeechRequest} = require('../stubs/riva/proto/riva_tts_pb');
 const {AudioEncoding} = require('../stubs/riva/proto/riva_audio_pb');
 const debug = require('debug')('jambonz:realtimedb-helpers');
-const EXPIRES = (process.env.JAMBONES_TTS_CACHE_DURATION_MINS || 24 * 60) * 60; // cache tts for 24 hours
+const EXPIRES = (process.env.JAMBONES_TTS_CACHE_DURATION_MINS || 4 * 60) * 60; // cache tts for 4 hours
 const TMP_FOLDER = '/tmp';
 
+
+const trimTrailingSilence = (buffer) => {
+  assert.ok(buffer instanceof Buffer, 'trimTrailingSilence - argument is not a Buffer');
+
+  let offset = buffer.length;
+  while (offset > 0) {
+    // Get 16-bit value from the buffer (read in reverse)
+    const value = buffer.readUInt16BE(offset - 2);
+    if (value !== 0) {
+      break;
+    }
+    offset -= 2;
+  }
+
+  // Trim the silence from the end
+  return offset === buffer.length ? buffer : buffer.subarray(0, offset);
+};
+
 /**
  * Synthesize speech to an mp3 file, and also cache the generated speech
  * in redis (base64 format) for 24 hours so as to avoid unnecessarily paying
@@ -104,7 +122,12 @@ async function synthAudio(client, logger, stats, { account_sid,
     text
   });
   let filePath;
-  if (['nuance', 'nvidia'].includes(vendor)) {
+  if (['nuance', 'nvidia'].includes(vendor) ||
+    (
+      process.env.JAMBONES_TTS_TRIM_SILENCE &&
+      ['microsoft', 'azure'].includes(vendor)
+    )
+  ) {
     filePath = `${TMP_FOLDER}/${key.replace('tts:', `tts-${salt || ''}`)}.r8`;
   }
   else filePath = `${TMP_FOLDER}/${key.replace('tts:', `tts-${salt || ''}`)}.mp3`;
@@ -284,6 +307,7 @@ const synthMicrosoft = async(logger, {
 }) => {
   try {
     const {api_key: apiKey, region, use_custom_tts, custom_tts_endpoint} = credentials;
+    const trimSilence = filePath.endsWith('.r8');
     let content = text;
     const speechConfig = SpeechConfig.fromSubscription(apiKey, region);
     speechConfig.speechSynthesisLanguage = language;
@@ -297,7 +321,9 @@ const synthMicrosoft = async(logger, {
        */
       if (!content.startsWith('<speak')) content = `<speak>${text}</speak>`;
     }
-    speechConfig.speechSynthesisOutputFormat = SpeechSynthesisOutputFormat.Audio16Khz32KBitRateMonoMp3;
+    speechConfig.speechSynthesisOutputFormat = trimSilence ?
+      SpeechSynthesisOutputFormat.Raw8Khz16BitMonoPcm :
+      SpeechSynthesisOutputFormat.Audio16Khz32KBitRateMonoMp3;
     const synthesizer = new SpeechSynthesizer(speechConfig);
 
     if (content.startsWith('<speak>')) {
@@ -323,7 +349,9 @@ const synthMicrosoft = async(logger, {
               reject(cancellation.errorDetails);
               break;
             case ResultReason.SynthesizingAudioCompleted:
-              resolve(Buffer.from(result.audioData));
+              let buffer = Buffer.from(result.audioData);
+              if (trimSilence) buffer = trimTrailingSilence(buffer);
+              resolve(buffer);
               synthesizer.close();
               stats.increment('tts.count', ['vendor:microsoft', 'accepted:yes']);
               break;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/speech-utils",
-  "version": "0.0.14",
+  "version": "0.0.16",
   "description": "TTS-related speech utilities for jambonz",
   "main": "index.js",
   "author": "Dave Horton",
@@ -24,7 +24,7 @@
   },
   "homepage": "https://github.com/jambonz/speech-utils#readme",
   "dependencies": {
-    "@aws-sdk/client-polly": "^3.303.0",
+    "@aws-sdk/client-polly": "^3.347.1",
     "@google-cloud/text-to-speech": "^4.2.1",
     "@grpc/grpc-js": "^1.8.13",
     "bent": "^7.3.12",