Merge pull request #1173 from omnivore-app/fix/speech

fix/speech

packages/api/src/routers/article_router.ts

@@ -29,7 +29,7 @@ interface SpeechInput {
   secondaryVoice?: string
   priority?: 'low' | 'high'
 }
-const outputFormats = ['mp3', 'speech-marks', 'speech-file']
+const outputFormats = ['mp3', 'speech-marks', 'speech']
 const logger = buildLogger('app.dispatch')
 
 export function articleRouter() {
@@ -102,7 +102,7 @@ export function articleRouter() {
         },
       })
 
-      if (outputFormat === 'speech-file') {
+      if (outputFormat === 'speech') {
         const page = await getPageById(articleId)
         if (!page) {
           return res.status(404).send('Page not found')
@@ -112,7 +112,7 @@ export function articleRouter() {
           secondaryVoice: secondaryVoice,
           language: page.language,
         })
-        return res.send(speechFile)
+        return res.send({ ...speechFile, pageId: articleId })
       }
 
       const existingSpeech = await getRepository(Speech).findOne({

packages/api/src/textToSpeech.d.ts

@@ -7,7 +7,7 @@ declare module '@omnivore/text-to-speech-handler' {
   export interface SSMLOptions {
     primaryVoice?: string
     secondaryVoice?: string
-    rate?: number
+    rate?: string
     language?: string
   }
 
@@ -21,7 +21,6 @@ declare module '@omnivore/text-to-speech-handler' {
 
   export interface SpeechFile {
     wordCount: number
-    averageWPM: number
     language: string
     defaultVoice: string
     utterances: Utterance[]

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

@@ -16,7 +16,6 @@ export interface Utterance {
 
 export interface SpeechFile {
   wordCount: number
-  averageWPM: number
   language: string
   defaultVoice: string
   utterances: Utterance[]
@@ -33,14 +32,13 @@ export type SSMLItem = {
 export type SSMLOptions = {
   primaryVoice?: string
   secondaryVoice?: string
-  rate?: number
+  rate?: string
   language?: string
 }
 
-const WORDS_PER_MINUTE = 200
 const DEFAULT_LANGUAGE = 'en-US'
 const DEFAULT_VOICE = 'en-US-JennyNeural'
-const DEFAULT_RATE = 1.25
+const DEFAULT_RATE = '1.0'
 
 const ANCHOR_ELEMENTS_BLOCKED_ATTRIBUTES = [
   'omnivore-highlight-id',
@@ -300,14 +298,13 @@ export const htmlToSpeechFile = (
         wordOffset,
         node.nodeName === 'BLOCKQUOTE' ? options.secondaryVoice : undefined
       )
-      utterances.push(utterance)
+      utterance.wordCount > 0 && utterances.push(utterance)
       wordOffset += utterance.wordCount
     }
   }
 
   return {
     wordCount: wordOffset,
-    averageWPM: WORDS_PER_MINUTE,
     language: options.language || DEFAULT_LANGUAGE,
     defaultVoice: options.primaryVoice || DEFAULT_VOICE,
     utterances,

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

@@ -13,7 +13,7 @@ import { htmlToSpeechFile } from './htmlToSsml'
 
 interface UtteranceInput {
   voice?: string
-  rate?: number
+  rate?: string
   language?: string
   text: string
   idx: string
@@ -24,7 +24,7 @@ interface HTMLInput {
   text: string
   voice?: string
   language?: string
-  rate?: number
+  rate?: string
   complimentaryVoice?: string
   bucket: string
 }
@@ -160,9 +160,6 @@ export const textToSpeechStreamingHandler = Sentry.GCPFunction.wrapHttpFunction(
 
     try {
       const utteranceInput = req.body as UtteranceInput
-      if (!utteranceInput.text) {
-        return res.status(400).send({ errorCode: 'INVALID_DATA' })
-      }
       const input: TextToSpeechInput = {
         ...utteranceInput,
         textType: 'utterance',

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

@@ -8,13 +8,14 @@ import {
   SpeechSynthesizer,
 } from 'microsoft-cognitiveservices-speech-sdk'
 import { endSsml, htmlToSsmlItems, ssmlItemText, startSsml } from './htmlToSsml'
+import * as _ from 'underscore'
 
 export interface TextToSpeechInput {
   text: string
   voice?: string
   language?: string
   textType?: 'html' | 'utterance'
-  rate?: number
+  rate?: string
   secondaryVoice?: string
   audioStream?: NodeJS.ReadWriteStream
 }
@@ -150,7 +151,7 @@ export const synthesizeTextToSpeech = async (
     // for utterance
     const start = startSsml(ssmlOptions)
     wordOffset = -start.length
-    const ssml = `${start}${input.text}${endSsml()}`
+    const ssml = `${start}${_.escape(input.text)}${endSsml()}`
     const result = await speakSsmlAsyncPromise(ssml)
     return {
       audioData: Buffer.from(result.audioData),

packages/text-to-speech/test/htmlToSsml.test.ts

@@ -7,7 +7,7 @@ describe('htmlToSsmlItems', () => {
     primaryVoice: 'test-primary',
     secondaryVoice: 'test-secondary',
     language: 'en-US',
-    rate: 1,
+    rate: '1.0',
   }
 
   describe('a simple html file', () => {