Transcribing Radio Broadcasts With Node.js

There are so many uses for Deepgram's live transcription service - from captioning meetings and events to creating home assistance and supporting call center operators by picking up on keywords.

Today, you'll use the Deepgram JavaScript SDK to provide live transcriptions to live radio broadcasts and store spoken words in a file that can then be further analyzed.

You will need a Deepgram API Key - get one here.

Create a new directory, open it on a code editor, and navigate to it in your terminal. Initialize a new Node.js project and install this project's dependencies:

npm init -y
npm install cross-fetch @deepgram/sdk

cross-fetch is used to make HTTP requests more straightforward in your Node.js projects. Alternatives include axios, got, and httpie - use whatever works for you or the default http library in Node.js, which requires no dependencies.

Create an index.js file and open it in your code editor. Initialize the project dependencies:

const fetch = require('cross-fetch')
const { Deepgram } = require('@deepgram/sdk')
const fs = require('fs')

Initialize the Deepgram JavaScript SDK, and create a new live transcription session:

const deepgram = new Deepgram('YOUR_DEEPGRAM_API_KEY')
const deepgramLive = deepgram.transcription.live({
  punctuate: true,
  tier: 'enhanced'
})

Two features are used in this session - punctuation and tier. Read more about Deepgram features such as redaction, diarization, and language.

Make sure you have a direct audio stream for the radio station. A good way of testing this is to open the URL in a browser - you should see just the built-in browser audio player without an accompanying web page.

Here are a few URLs for you to try:

• BBC World Service (works outside the UK): http://stream.live.vc.bbcmedia.co.uk/bbc_world_service
• France Inter: https://direct.franceinter.fr/live/franceinter-midfi.mp3

If you use the French channel, be sure to add language: fr to your Deepgram session options.

const url = 'http://stream.live.vc.bbcmedia.co.uk/bbc_world_service'

fetch(url).then(r => r.body).then(res => {
  res.on('readable', () => {
    const data = res.read()
    console.log(data)
  })
})

Run your code with node index.js, leave it running for a couple of seconds, and stop it with ctrl+c. You should see a bunch of buffers logged to your console.

This is what you want to see - these buffers of audio data can be sent directly to Deepgram.

Replace console.log(data) with the following to send the buffers to Deepgram if the connection is still open:

if(deepgramLive.getReadyState() === 1) {
    deepgramLive.send(data)
}

At the bottom of index.js, below all other code, add this code to listen for returned transcripts:

deepgramLive.addListener('transcriptReceived', (message) => {
  const data = JSON.parse(message)
  const transcript = data.channel.alternatives[0].transcript
  if(transcript) {
    console.log(transcript)
  }
})

Rerun your code, and you should see transcripts in your terminal.

To save these transcripts to a file, you must first create a write stream and then write content to it. At the top of your file, just below your require statements, create the stream:

const stream = fs.createWriteStream('output.txt', { flags: 'a' })

The a flag will open the file specifically for appending new data. If it does not exist, it will be automatically created.

Replace console.log(transcript) with the following:

stream.write(transcript + ' ')

This will add the new transcript to the end of the existing file, ensuring there is a space between each item.

Run your code again, wait a few seconds, and then stop it. Take a look at the new output.txt file, and you should see a big block of text which can then be stored in a database for compliance or further analysis.

The full code is here:

const fetch = require('cross-fetch')
const { Deepgram } = require('@deepgram/sdk')
const fs = require('fs')
const stream = fs.createWriteStream('output.txt', { flags:'a' })

const deepgram = new Deepgram(deepgramApiKey)
const deepgramLive = deepgram.transcription.live({
  punctuate: true,
  tier: 'enhanced'
})

const url = 'http://stream.live.vc.bbcmedia.co.uk/bbc_world_service'

fetch(url).then(r => r.body).then(res => {
  res.on('readable', () => {
    const data = res.read()
    if(deepgramLive.getReadyState() === 1) {
      deepgramLive.send(data)
    }
  })
})

deepgramLive.addListener('transcriptReceived', (message) => {
  const data = JSON.parse(message)
  const transcript = data.channel.alternatives[0].transcript
  if(transcript) {
    stream.write(transcript + ' ')
  }
})

If you have any questions, please feel free to reach out to us over email (devrel@deepgram.com) or via Twitter (@DeepgramDevs).

If you have any feedback about this post, or anything else around Deepgram, we'd love to hear from you. Please let us know in our GitHub discussions .