Tavus introduced Griffin, the first Human Interaction Model (HIM).
48% of people in live one-minute video calls thought it was human. Prior systems were under 3%. It ranks #1 on NVIDIA’s full-duplex video benchmark.
It generates the full frame, not just a face: plays Simon Says by watching and moving its hands, laughs and nods mid-sentence, folds objects you hold up into the conversation, and coaches Rubik’s cube or soldering from what it sees.
Real-time video that listens, watches, and talks at once.
@tavus