From 4b42d013cafd0f597a02c4ac571138bb8568f84d Mon Sep 17 00:00:00 2001
From: Hongbo Wu <hongbo@omnivore.app>
Date: Fri, 12 Aug 2022 11:49:11 +0800
Subject: [PATCH] Convert text to OGG file

---
 packages/api/src/utils/textToSpeech.ts       | 8 ++++----
 packages/api/test/utils/textToSpeech.test.ts | 4 ++--
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/packages/api/src/utils/textToSpeech.ts b/packages/api/src/utils/textToSpeech.ts
index 34e2e860b..5d3f40eca 100644
--- a/packages/api/src/utils/textToSpeech.ts
+++ b/packages/api/src/utils/textToSpeech.ts
@@ -26,7 +26,7 @@ export const createAudio = async (
 ): Promise<Buffer> => {
   const { text, voice, textType, engine } = input
   const params: SynthesizeSpeechInput = {
-    OutputFormat: 'mp3',
+    OutputFormat: 'ogg_vorbis',
     Text: text,
     TextType: textType || 'text',
     VoiceId: voice || 'Joanna',
@@ -51,7 +51,7 @@ export const createSpeechMarks = async (
     TextType: textType || 'text',
     VoiceId: voice || 'Joanna',
     Engine: engine || 'neural',
-    SpeechMarkTypes: ['sentence', 'word'],
+    SpeechMarkTypes: ['sentence'],
   }
   try {
     const data = await client.synthesizeSpeech(params).promise()
@@ -68,11 +68,11 @@ export const createAudioWithSpeechMarks = async (
   try {
     const audio = await createAudio(input)
     // upload audio to google cloud storage
-    const filePath = `speech/${input.id}.mp3`
+    const filePath = `speech/${input.id}.ogg`
 
     logger.info('start uploading...', { filePath })
     await uploadToBucket(filePath, audio, {
-      contentType: 'audio/mpeg',
+      contentType: 'audio/ogg',
       public: true,
     })
 
diff --git a/packages/api/test/utils/textToSpeech.test.ts b/packages/api/test/utils/textToSpeech.test.ts
index 62a3e53c2..0dc698c33 100644
--- a/packages/api/test/utils/textToSpeech.test.ts
+++ b/packages/api/test/utils/textToSpeech.test.ts
@@ -11,9 +11,9 @@ describe('textToSpeech', () => {
     it('should create an audio file with speech marks', async () => {
       const input: TextToSpeechInput = {
         id: generateFakeUuid(),
-        title: 'Hello World',
-        text: 'Hello World',
+        text: '<speak><mark name="readability-page-1"/>The rumor mill suggests that Google may be looking to kill off its game streaming platform, Stadia, for good before the end of the year.</speak>',
         engine: 'standard',
+        textType: 'ssml',
       }
       const output = await createAudioWithSpeechMarks(input)
       expect(output.audioUrl).to.be.a('string')