Transcription Overview
Start generating transcripts for your recorded meetings.
One of the main use cases of meeting recording is to get an accurate, speaker-attributed transcript. Recall handles the this process for you: it captures audio/video from meetings, routes it to your chosen speech-to-text provider, applies speaker attribution, and delivers the transcript in a consistent format. This can be done in real time or after the call ends. You choose the speech to text provider: Recall’s built-in engine, a third party, or the meeting platform’s captions.
Transcription must be enabledTranscripts are not generated for meetings by default.
To generate a transcript, you must be configure a transcription method.
Quickstart
If you're looking for the fastest way to get started with transcription, use these guides:
- Bots: Follow the Recall.ai Transcription guide
- Desktop Recording SDK: Follow the Desktop SDK Real-Time Transcription guide
Transcription Providers
Recall allows you to choose between several different transcription options, depending on your use case:
Recall.ai Transcription : Transcription directly through Recall.ai
Third-Party Transcription: Through one of our transcription partners
Meeting Caption Transcription: Using native meeting platform captions
Recall will return transcripts in the same API response format no matter which of these you use, so feel free to test each of them out.
Which provider should I use?
Transcription Method | Pros | Cons |
---|---|---|
Recall.ai Transcription |
|
|
Third-Party Transcription |
|
|
Meeting Captions |
|
|
Third-Party Transcription Providers
The following is a list of officially-supported third-party transcription providers. While we support real-time and async transcription on both meeting bots and the Desktop Recording SDK, some providers are not available for real-time on the Desktop SDK yet.
Transcription Provider | Async Transcription (all) | Realtime Transcription (Bots) | Realtime Transcription (Desktop SDK) |
---|---|---|---|
Assembly AI | ✅ | ✅ | ✅ |
Deepgram | ✅ | ✅ | ✅ |
Gladia | ✅ | ✅ | ✅ |
AWS Transcribe | ❌ | ✅ | ❌ |
Rev | ✅ | ✅ | ❌ |
Speechmatics | ✅ | ✅ | ❌ |
Transcription Methods
Recall supports transcription both in real time and after the call ends. Choose real-time when you need the transcript as soon as possible, and choose async for best final transcript quality.
Real-time transcription
Real‑time transcription is delivered continuously during the call via webhook or websocket. See the following guides for how to implement real-time transcription for both bots and the Desktop Recording SDK:
Desktop SDK Real-time Transcription
Async transcription
Post-call transcription yields the highest overall transcription quality, so you should prefer this method unless you need the transcript data in real time. The implementation for async transcription is the same for both bots and the Desktop Recording SDK, so you can follow this guide:
Diarization
Diarization is the process of determining who said what in a transcript. Recall automatically diarizes transcripts using a method called speaker timeline diarization, but there are other options available. Read more about this here:
Updated about 4 hours ago