Detect word boundary only in speech marks

packages/text-to-speech/src/textToSpeech.ts

@@ -1,10 +1,8 @@
 import {
   CancellationDetails,
   CancellationReason,
-  PropertyId,
   ResultReason,
   SpeechConfig,
-  SpeechSynthesisBoundaryType,
   SpeechSynthesisOutputFormat,
   SpeechSynthesisResult,
   SpeechSynthesizer,
@@ -32,7 +30,7 @@ export interface SpeechMark {
   start?: number
   length?: number
   word: string
-  type: 'word' | 'bookmark' | 'punctuation' | 'sentence'
+  type: 'word' | 'bookmark'
 }
 
 export const synthesizeTextToSpeech = async (
@@ -49,17 +47,11 @@ export const synthesizeTextToSpeech = async (
   )
   speechConfig.speechSynthesisOutputFormat =
     SpeechSynthesisOutputFormat.Audio16Khz32KBitRateMonoMp3
-  // Required for sentence-level WordBoundary events
-  speechConfig.setProperty(
-    PropertyId.SpeechServiceResponse_RequestSentenceBoundary,
-    'true'
-  )
 
   // Create the speech synthesizer.
   const synthesizer = new SpeechSynthesizer(speechConfig)
   const speechMarks: SpeechMark[] = []
   let timeOffset = 0
-  // let wordOffset = 0
 
   synthesizer.synthesizing = function (s, e) {
     // convert arrayBuffer to stream and write to stream
@@ -94,14 +86,13 @@ export const synthesizeTextToSpeech = async (
 
   // The unit of e.audioOffset is tick (1 tick = 100 nanoseconds), divide by 10,000 to convert to milliseconds.
   synthesizer.wordBoundary = (s, e) => {
-    e.boundaryType === SpeechSynthesisBoundaryType.Sentence &&
-      speechMarks.push({
-        word: e.text,
-        time: (timeOffset + e.audioOffset) / 10000,
-        start: e.textOffset,
-        length: e.text.length,
-        type: 'sentence',
-      })
+    speechMarks.push({
+      word: e.text,
+      time: (timeOffset + e.audioOffset) / 10000,
+      start: e.textOffset,
+      length: e.text.length,
+      type: 'word',
+    })
   }
 
   synthesizer.bookmarkReached = (s, e) => {
@@ -150,8 +141,6 @@ export const synthesizeTextToSpeech = async (
     const startSsmlTag = startSsml(ssmlOptions)
     const text = _.escape(input.text)
     const ssml = `${startSsmlTag}${text}${endSsml()}`
-    // set the text offset to be the end of SSML start tag
-    // wordOffset -= startSsmlTag.length
     const result = await speakSsmlAsyncPromise(ssml)
     if (result.reason === ResultReason.Canceled) {
       throw new Error(result.errorDetails)