Getting Started

This module makes it easy to create real-time speech-to-text (s2t) transcription from audio data.

Installation

npm i --save @bottlenose/rxtranscribe

yarn add @bottlenose/rxtranscribe

Configure an AWS account

To use this package, you'll need a properly configured AWS account. The account needs to be configured to allow real-time streaming via AWS Transcribe.

By default the transcribe operator in @bottlenose/rxtranscribe looks for AWS credentials and configuration stored in environment variables: AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY and AWS_REGION. You can also pass them in as optional parameters. (See the transcribe operator docs for up-to-date information.)

Using the operator

Once you have AWS credentials setup, you can transcribe audio streams in real-time:

import {fromFile} from '@bottlenose/rxfs';
import {transcribe} from '@bottlenose/rxtranscribe';

const transcribeAudioFromFile = function transcribeAudioFromFile({
  inputFilePath = path.resolve('sample-audio.mp3'),
}) {
  const transcriptionOptions = {};
  const mp3Chunk$ = fromFile({filePath: inputFilePath});
  const transcription$ = mp3Chunk$.pipe(
    transcribe(transcriptionOptions)
  );
  return transcription$;
};

export default transcribeAudio;

The example above uses data streamed from a local file but the transcribe operator can be performed on any input observable that contains input data acceptable to AWS Transcribe's real-time streaming API (generally .wav files like .mp3).