OpenAI Whisper for Node

Locally converts recorded audio, like WAV files and MP3s, to text using OpenAI Whisper.

Prerequisites

Like other Node wrappers around OpenAI Whisper, this package requires a few tools to work.

Python

Download from the official Python website. For best results, make sure to add Python to the path.

FFmpeg

For audio processing.

On Windows via Chocolatey:

choco install ffmpeg

On Debian-based Linux:

sudo apt-get install ffmpeg

OpenAI Whisper

Probably the easiest way to install Whisper is using the Python package manager, pip.

pip install -U openai-whisper

To test your installation:

whisper --help

Installation

npm install @cityssm/whisper-speech-to-text

Usage

import speechToText from '@cityssm/whisper-speech-to-text'

const basicTranscription = await speechToText('path/to/audioFile.mp3')

const transcriptionWithOptions = await speechToText('path/to/audioFile.wav', {
  whisperPath: 'path/to/whisper',
  model: 'large',
  language: 'en'
})