Transcribe audio

Send an audio file and collect its transcript.

This page describes the v0 API from v0.5.1. Use v0 package builds with these examples. Start a compatible service at the local address shown in the example, or replace that address with your service root. Make sure that the service accepts the model ID and request protocol shown below. Use Node.js with openAsBlob support. Supply an audio file at the example path.

npm i @xsai/generate-transcription@0.5.1

Examples

Basic

const { text } = await generateTranscription({
  apiKey: '',
  baseURL: 'http://localhost:8000/v1/',
  file: await openAsBlob('./test/fixtures/basic.wav', { type: 'audio/wav' }),
  fileName: 'basic.wav',
  language: 'en',
  model: 'deepdml/faster-whisper-large-v3-turbo-ct2',
})

Verbose + Segments

const { duration, language, segments, text } = await generateTranscription({ 
  apiKey: '',
  baseURL: 'http://localhost:8000/v1/',
  file: await openAsBlob('./test/fixtures/basic.wav', { type: 'audio/wav' }),
  fileName: 'basic.wav',
  language: 'en',
  model: 'deepdml/faster-whisper-large-v3-turbo-ct2',
  responseFormat: 'verbose_json', 
})

Verbose + Words

const { duration, language, text, words } = await generateTranscription({ 
  apiKey: '',
  baseURL: 'http://localhost:8000/v1/',
  file: await openAsBlob('./test/fixtures/basic.wav', { type: 'audio/wav' }),
  fileName: 'basic.wav',
  language: 'en',
  model: 'deepdml/faster-whisper-large-v3-turbo-ct2',
  responseFormat: 'verbose_json', 
  timestampGranularities: 'word', 
})

Result

The basic result contains text. Verbose results can also contain duration, language, segments, or words.