You can also set GLADIA_API_URL. For live sessions, GLADIA_REGION (eu-west / us-west) is passed only on session creation (POST /v2/live).

Node.js / Browser (ESM)

import { GladiaClient } from '@gladiaio/sdk'

const gladiaClient = new GladiaClient({
  apiKey: 'your-api-key',
})

Node.js (CommonJS)

const { GladiaClient } = require('@gladiaio/sdk')

const gladiaClient = new GladiaClient({
  apiKey: 'your-api-key',
})

Browser (script tag)

<script src="https://unpkg.com/@gladiaio/sdk"></script>

<script>
  const gladiaClient = new Gladia.GladiaClient({
    apiKey: 'your-api-key',
  })
</script>

Pre-recorded transcription

transcribe() accepts a file path (Node), http(s) URL, File, or Blob. It uploads when needed, then polls until the job completes.

import { GladiaClient } from '@gladiaio/sdk'
import 'dotenv/config'

const gladiaClient = new GladiaClient()
const audioPath = '../data/online-meeting-example.mp4'

const result = await gladiaClient.preRecorded().transcribe(audioPath, {
  language_config: { languages: ['en'] },
})

console.log(result.result?.transcription?.full_transcript ?? '')

See all supported languages here !

Pass the options argument to enable features from Audio intelligence such as diarization, translation, PII redaction, and much more.

Async pre-recorded

If your runtime has no top-level await, wrap calls in an async function:

async function main() {
  const gladiaClient = new GladiaClient()
  const result = await gladiaClient.preRecorded().transcribe('./audio.mp3')
  console.log(result.result?.transcription?.full_transcript ?? '')
}

main().catch(console.error)

Live transcription

const liveSession = gladiaClient.liveV2().startSession({
  model: 'solaria-1',
  encoding: 'wav/pcm',
  sample_rate: 16000,
  bit_depth: 16,
  channels: 1,
  language_config: {
    languages: ['en'],
  },
  messages_config: {
    receive_partial_transcripts: true,
  },
})

liveSession.on('message', (message) => {
  if (message.type === 'transcript') {
    console.log(`${message.data.is_final ? 'F' : 'P'} | ${message.data.utterance.text.trim()}`)
  }
})

liveSession.once('started', () => {
  console.log(`Session ${liveSession.sessionId} started`)
})

liveSession.on('error', (err) => {
  console.error('An error occurred during live session:', err)
})

liveSession.once('ended', () => {
  console.log(`Session ${liveSession.sessionId} ended`)
})

liveSession.sendAudio(/* <audio_chunk> */)
liveSession.stopRecording()

See all the supported languages here !

Pass the options argument to enable features from Audio intelligence such as diarization, translation, PII redaction, and much more.

Waiting for the session to finish

To await shutdown after stopRecording(), wait on the ended event:

import { GladiaClient } from '@gladiaio/sdk'

async function runLive() {
  const gladiaClient = new GladiaClient({ apiKey: 'your-api-key' })
  const liveSession = gladiaClient.liveV2().startSession({
    language_config: { languages: ['en'] },
  })

  const ended = new Promise<void>((resolve) => {
    liveSession.once('ended', () => resolve())
  })

  liveSession.stopRecording()
  await ended
}

runLive().catch(console.error)

When you need the session id as soon as the backend has created the session:

const sessionId = await liveSession.getSessionId()

@gladiaio/sdk

Gladia JavaScript SDK

Requirements

Installation

Usage