Project aimed at utilizing Azure AI Text to Speech service to build a text reader for news sites, Wikipedia etc.
What is Azure Speech Service?
The Speech service provides speech to text and text to speech capabilities with a Speech resource. You can transcribe speech to text with high accuracy, produce natural-sounding text to speech voices, translate spoken audio, and use speaker recognition during conversations.
Front is a chrome extension containing a popup.html to displayer the audio player and 3 layers of scripts:
i. content.js - to collect article text from page. ii. background.js - service worker to handle messages from content.js and popup.js. iii. popup.js - to fetch audio from backend and serve audio file to popup.html
- Currently popup.js retriggers fetch everytime it is closed and opened. We need to identify a way to persist the audio as long as the tab is open to reduce the calls to the backend.
- Create a more robust scraper for the frontend. Currently it is simply looking for the first article tag.
Backend is built on Express.js to serve requests from the front end. Currently, there is only one service - it accepts text and responds with a Readable Stream of WAV audio.
-
OPTIONS request for "/audio" handles pre-flight requests from chrome extension to prevent any CORS exceptions.
-
POST request for "/audio" calls the generateAudioFile function.
i. generateAudioFile creates a new instance of Azure speech configuration:
const speechConfig = sdk.SpeechConfig.fromSubscription(
process.env.SPEECH_KEY,
process.env.SPEECH_REGION
);ii. A new instance of SpeechSynthesizer is created to perform speech synthesis audio.
var synthesizer = new sdk.SpeechSynthesizer(speechConfig);iv. speakTextAsync produces an ArrayBuffer object called audioData. The function takes an ArrayBuffer as input and converts it to Buffer using Buffer.from(arrayBuffer). Then, it creates a Readable stream, pushes the Buffer data to the stream using stream.push(buffer), and finally signals the end of the stream using stream.push(null).