0.1.16

fix playht has stream and cached audio differrent quality

.github/workflows/ci.yml

@@ -13,6 +13,11 @@ jobs:
         with:
           node-version: lts/*
       - run: npm install
+      - name: Install Docker Compose
+        run: |
+          sudo curl -L "https://github.com/docker/compose/releases/download/1.29.2/docker-compose-$(uname -s)-$(uname -m)" -o /usr/local/bin/docker-compose
+          sudo chmod +x /usr/local/bin/docker-compose
+          docker-compose --version
       - run: npm run jslint
       - run: sudo apt update && sudo apt install -y squid
       - run: sudo cp test/squid.conf /etc/squid/squid.conf

lib/config.js

@@ -1,6 +1,7 @@
 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_HTTP_PROXY_IP = process.env.JAMBONES_HTTP_PROXY_IP;
 const JAMBONES_HTTP_PROXY_PORT = process.env.JAMBONES_HTTP_PROXY_PORT;
@@ -18,6 +19,7 @@ module.exports = {
   JAMBONES_HTTP_PROXY_IP,
   JAMBONES_HTTP_PROXY_PORT,
   JAMBONES_TTS_CACHE_DURATION_MINS,
+  JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
   TMP_FOLDER,
   HTTP_TIMEOUT
 };

lib/synth-audio.js

@@ -44,6 +44,7 @@ const {
   JAMBONES_HTTP_PROXY_IP,
   JAMBONES_HTTP_PROXY_PORT,
   JAMBONES_TTS_CACHE_DURATION_MINS,
+  JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
 } = require('./config');
 const EXPIRES = JAMBONES_TTS_CACHE_DURATION_MINS;
 const OpenAI = require('openai');
@@ -86,7 +87,7 @@ const trimTrailingSilence = (buffer) => {
  */
 async function synthAudio(client, createHash, retrieveHash, logger, stats, { account_sid,
   vendor, language, voice, gender, text, engine, salt, model, credentials, deploymentId,
-  disableTtsCache, renderForCaching, disableTtsStreaming, options
+  disableTtsCache, renderForCaching = false, disableTtsStreaming, options
 }) {
   let audioBuffer;
   let servedFromCache = false;
@@ -157,20 +158,42 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
     }
 
   }
+
   const key = makeSynthKey({
     account_sid,
     vendor,
     language: language || '',
     voice: voice || deploymentId,
     engine,
-    text
+    text,
+    renderForCaching
   });
   let filePath;
-  filePath = makeFilePath(vendor, key, salt);
+  filePath = makeFilePath({vendor, key, salt, renderForCaching});
   debug(`synth key is ${key}`);
   let cached;
   if (!disableTtsCache) {
     cached = await client.get(key);
+    /**
+    *  If we are using tts streaming and also precaching audio, audio could have been cached by streaming (r8)
+    * or here in speech-utils due to precaching (mp3), so we need to check for both keys.
+    */
+    if (!cached && JAMBONES_EAGERLY_PRE_CACHE_AUDIO) {
+      const preCachekey = makeSynthKey({
+        account_sid,
+        vendor,
+        language: language || '',
+        voice: voice || deploymentId,
+        engine,
+        text,
+        renderForCaching: true
+      });
+      cached = await client.get(preCachekey);
+      if (cached) {
+        // Precache audio is available update filpath with precache file extension.
+        filePath = makeFilePath({vendor, key, salt, renderForCaching: true});
+      }
+    }
   }
   if (cached) {
     // found in cache - extend the expiry and use it

lib/utils.js

@@ -16,29 +16,31 @@ const debug = require('debug')('jambonz:realtimedb-helpers');
  */
 //const nuanceClientMap = new Map();
 
-function makeSynthKey({account_sid = '', vendor, language, voice, engine = '', text}) {
+function makeSynthKey({
+  account_sid = '', vendor, language, voice, engine = '', text,
+  renderForCaching = false}) {
   const hash = crypto.createHash('sha1');
   hash.update(`${language}:${vendor}:${voice}:${engine}:${text}`);
   const hexHashKey = hash.digest('hex');
   const accountKey = account_sid ? `:${account_sid}` : '';
-  const namespace = vendor.startsWith('custom') ? vendor : getFileExtension(vendor);
+  const namespace = vendor.startsWith('custom') ? vendor : getFileExtension({vendor, renderForCaching});
   const key = `tts${accountKey}:${namespace}:${hexHashKey}`;
   return key;
 }
 
-function makeFilePath(vendor, key, salt = '') {
-  const extension = getFileExtension(vendor);
+function makeFilePath({vendor, key, salt = '', renderForCaching = false}) {
+  const extension = getFileExtension({vendor, renderForCaching});
   return `${TMP_FOLDER}/${key.replace('tts:', `tts-${salt}`)}.${extension}`;
 }
 
-function getFileExtension(vendor) {
+function getFileExtension({vendor, renderForCaching = false}) {
   const mp3Extension = 'mp3';
   const r8Extension = 'r8';
 
   switch (vendor) {
     case 'azure':
     case 'microsoft':
-      if (!JAMBONES_DISABLE_TTS_STREAMING || JAMBONES_TTS_TRIM_SILENCE) {
+      if (!renderForCaching && !JAMBONES_DISABLE_TTS_STREAMING || JAMBONES_TTS_TRIM_SILENCE) {
         return r8Extension;
       } else {
         return mp3Extension;
@@ -46,16 +48,25 @@ function getFileExtension(vendor) {
     case 'deepgram':
     case 'elevenlabs':
     case 'rimlabs':
-      if (!JAMBONES_DISABLE_TTS_STREAMING) {
-        return r8Extension;
-      } else {
+    case 'playht':
+      if (renderForCaching || JAMBONES_DISABLE_TTS_STREAMING) {
         return mp3Extension;
+      } else {
+        return r8Extension;
       }
     case 'nuance':
     case 'nvidia':
     case 'verbio':
       return r8Extension;
     default:
+      // If vendor is custom
+      if (vendor.startsWith('custom')) {
+        if (renderForCaching || JAMBONES_DISABLE_TTS_STREAMING) {
+          return mp3Extension;
+        } else {
+          return r8Extension;
+        }
+      }
       return mp3Extension;
   }
 }

package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "@jambonz/speech-utils",
-  "version": "0.1.13",
+  "version": "0.1.16",
   "lockfileVersion": 2,
   "requires": true,
   "packages": {
     "": {
       "name": "@jambonz/speech-utils",
-      "version": "0.1.13",
+      "version": "0.1.16",
       "license": "MIT",
       "dependencies": {
         "@aws-sdk/client-polly": "^3.496.0",

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/speech-utils",
-  "version": "0.1.13",
+  "version": "0.1.16",
   "description": "TTS-related speech utilities for jambonz",
   "main": "index.js",
   "author": "Dave Horton",