Unknown · since 2026-10-01 · 2× · last seen Oct 09, 2026
40
Momentum
Griffin is a Human Interaction Model from Tavus, introduced on October 1, 2026. It is a full-duplex video-to-video system that listens, watches and responds at the same time, instead of chaining speech, language model, voice and avatar in sequence. Only the research preview Griffin-Lite is available, to selected testers; the full model is not released.
Momentum trend
11.07.09.10.
Features
| Real-Time Streaming | Yes, full-duplex: listens, watches and speaks simultaneously; output in 320 ms chunks |
| Latency | Avg. 0.43 s audio-to-video on H100 GPUs (Tavus measurement) |
| Platform | Not on the regular Tavus platform; access only via request form for selected testers |
| Price | Not published |
| Release Date | October 1, 2026 (announced); Griffin-Lite as research preview |
| Voice Cloning | Voice clone from about 10 seconds of audio (Griffin-Lite) |