From d9dc701de80f6c4a75d829dd1375a7fd56b3e67d Mon Sep 17 00:00:00 2001
From: Hongbo Wu <hongbo@omnivore.app>
Date: Wed, 7 Sep 2022 18:03:13 +0800
Subject: [PATCH] Stop counting the words in SSML tags

---
 packages/text-to-speech/src/textToSpeech.ts | 23 +++++++++++----------
 1 file changed, 12 insertions(+), 11 deletions(-)

diff --git a/packages/text-to-speech/src/textToSpeech.ts b/packages/text-to-speech/src/textToSpeech.ts
index 3bfe1be86..b7fef7368 100644
--- a/packages/text-to-speech/src/textToSpeech.ts
+++ b/packages/text-to-speech/src/textToSpeech.ts
@@ -13,7 +13,7 @@ export interface TextToSpeechInput {
   text: string
   voice?: string
   language?: string
-  textType?: 'html' | 'ssml' | 'utterance'
+  textType?: 'html' | 'utterance'
   rate?: number
   complimentaryVoice?: string
   audioStream?: NodeJS.ReadWriteStream
@@ -51,6 +51,7 @@ export const synthesizeTextToSpeech = async (
   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
@@ -93,7 +94,7 @@ export const synthesizeTextToSpeech = async (
     speechMarks.push({
       word: e.text,
       time: (timeOffset + e.audioOffset) / 10000,
-      start: e.textOffset,
+      start: wordOffset + e.textOffset,
       length: e.wordLength,
       type: 'word',
     })
@@ -142,15 +143,19 @@ export const synthesizeTextToSpeech = async (
         const result = await speakSsmlAsyncPromise(ssml)
         timeOffset = timeOffset + result.audioDuration
       }
-    } else {
-      // assemble ssml
-      const ssml = `${startSsml(null, ssmlOptions)}${input.text}${endSsml()}`
-      const result = await speakSsmlAsyncPromise(ssml)
       return {
-        audioData: Buffer.from(result.audioData),
         speechMarks,
       }
     }
+    // for utterance
+    const start = startSsml(ssmlOptions)
+    wordOffset = -start.length
+    const ssml = `${start}${input.text}${endSsml()}`
+    const result = await speakSsmlAsyncPromise(ssml)
+    return {
+      audioData: Buffer.from(result.audioData),
+      speechMarks,
+    }
   } catch (error) {
     console.error('synthesis error', error)
     throw error
@@ -160,8 +165,4 @@ export const synthesizeTextToSpeech = async (
     synthesizer.close()
     console.debug('synthesizer closed')
   }
-
-  return {
-    speechMarks,
-  }
 }