Overview
Predict when the user has finished speaking, on your servers with the VIVA SDK or as turn events from the Krisp Speech-to-Text API.
Turn-Taking predicts when the user has finished their turn, so your voice AI agent responds without awkward pauses and without cutting the user off. It replaces silence-based end-of-turn detection with a model that listens to how the user speaks.
Choose SDK or API
| VIVA SDK | Speech-to-Text API: STT for Agents | |
|---|---|---|
| Where it runs | On your servers, inside your audio pipeline (VIVA SDK) | In Krisp's cloud, as part of speech-to-text |
| What you get | An end-of-turn probability for every frame | A turn event on the transcript stream |
| Voice activity | From your own VAD | Built in |
| Tuning | Your probability threshold | turn.sensitivity, turn.max_silence, turn.silence_validation |
| Use it when | You run your own STT, or use a framework such as Pipecat | You also want Krisp's transcripts |
The Turn-Taking (VIVA SDK) page compares the two in more detail.
Key points
- Run Turn-Taking on audio cleaned by Voice Isolation. Background voices otherwise look like the user still speaking.
- The SDK model (TTv3,
krisp-viva-tp-v3) works with all languages. The official samples use 20 ms input frames. - Moving from TTv2? See Migrating to TTv3.
Updated 3 days ago
Did this page help you?