Realtime avatar model
Narvatar-1NEW
A living face streamed over one WebSocket. It listens while you talk, answers while you pause, and takes an interruption without losing the thread.
We simulate humans.
Nari Labs is an AI lab focused on multimodal, realtime, interactive intelligence.
Our models
Open-weight voice and avatar models you can run, fork, and talk to.
Dialogue TTS model
Dia
Two speakers, one pass. Dia renders dialogue with the timing of real conversation: laughter, sighs, overlaps and the pause before an honest answer.
6.8K
MONTHLY DOWNLOADS
2.9K
LIKES
Latest updates
All posts
Talk to the humans, too.
The avatar answers instantly. We're a little slower, but we read everything.
