# Speech Recognition

Real-time speech-to-text for live events from several leading engines, with the best one recommended for each language and a dictionary built from your event.

- **Languages**: 104
- **Engines**: 5
- **Latency**: 3s

Engines: Speechmatics, Deepgram, Gladia, AWS Transcribe, Microsoft MAI-Transcribe.

## What it does

### The right engine for each language

Line 21 recommends the best-performing engine for every language, and you can switch engines at any time.

### Your words, recognised

AI Context turns your scripts, slides and speaker names into a dictionary of words the engine should expect.

[AI Context](https://line-21.com/docs/ai-services/ai-context)

### Speakers and blended languages

Some engines tell speakers apart and follow conversations that switch between languages mid-sentence.

[AWS Transcribe](https://line-21.com/features/speech-recognition/aws-transcribe)

### Audio from anywhere

Feed audio from a browser microphone, the Line 21 desktop app, or an RTMP or SRT video stream.

[Audio feeds](https://line-21.com/docs/feeds/audio)

## Learn more

- [Speech recognition docs](https://line-21.com/docs/methods/asr)
- [Video feeds](https://line-21.com/docs/feeds/video)
- [Supported languages](https://line-21.com/languages)

Talk to us: [contact page](https://line-21.com/contact) or hello@line-21.com.