Live Streaming Audio Transcription

An overview of the Deepgram JavaScript SDK and Deepgram speech-to-text live streaming.

The Deepgram JavaScript SDK now works in both server and browser environments.

Installing the SDK

# Install the Deepgram JS SDK
# https://github.com/deepgram/deepgram-js-sdk

npm install @deepgram/sdk

Initialising the SDK

import { createClient } from "@deepgram/sdk";

const deepgram = createClient("DEEPGRAM_API_KEY");

Connecting to Deepgram

Connecting to Deepgram returns an instance of LiveClient.

const live = deepgram.listen.live({ model: "nova" });

Live Transcription Options

Transcription options are provided as an object to the live(...options) method. Each of these parameters are found as unique features on the Streaming API reference.

Events

Event's are available from the SDK as an enum.

import { LiveTranscriptionEvents } from "@deepgram/sdk";

Events sent by the SDK

The following events are returned by the SDK.

EventDescriptionData
LiveTranscriptionEvents.OpenThe web socket connection to Deepgram has been opened.Instance of LiveClient
LiveTranscriptionEvents.CloseThe web socket connection to Deepgram has been closed.WebSocket.CloseEvent
LiveTranscriptionEvents.TranscriptDeepgram has responded with a transcription result.Results
LiveTranscriptionEvents.MetadataDeepgram has responded with metadata.Metadata
LiveTranscriptionEvents.ErrorAn error occurred.Error
LiveTranscriptionEvents.WarningThe SDK has issued a warning to do with your request/usage.Warning

Listening to Events

The events occur as native JS EventListener events. Please note, .on() is an alias of addEventListener on an EventEmitter.

live.on(LiveTranscriptionEvents.Open, () => {
  live.on(LiveTranscriptionEvents.Transcript, (data) => {
    console.log(data);
  });
});

Functions

An instance of LiveClient has the following functionality.

Sending Data

You can send raw audio data or text to the web socket without a content type. We will determine the appropriate encoding from the data.

 live.send(data); 

Keep Alive

You can send Deepgram a keep alive if you intend for deliberate moments of non-billable silence while the socket remains open. The connection will close after 12 seconds if no data is sent, but you will still be billed for an active connection if you're sending data, even if no audible words are included in the data.

live.keepAlive()

📘

Read more about KeepAlive in this comprehensive guide.

Get Ready State

The getReadyState function returns the ready state of the web socket connection to Deepgram.

live.getReadyState();

Toggle Features

The configure function gives the ability to toggle different features. Currently, only the numerals feature is supported. This function is a shorthand way of sending the following data to the web socket connection:

live.configure({numerals: true}); // enable numerals

Closing the Connection

The requestClose function closes the web socket connection to Deepgram.

live.requestClose();