Tavus has introduced Griffin, the first Human Interaction Model designed for real-time face-to-face conversations. In a recent company study, nearly half of test subjects mistook the AI video avatar for a real person during a one-minute video call. Previous conversational systems maxed out at just two percent. Furthermore, in an independent Nvidia test measuring conversational human likeness, Griffin scored 3.83 points out of four, coming very close to the 3.92 score achieved by actual humans while vastly outperforming prior AI models.
The advanced model processes speech, facial expressions, tone of voice, gestures, and pauses simultaneously while both receiving and generating video. A preview version called Griffin-Lite is currently available to select testers as a research preview, and a more capable iteration will follow once safety concerns are addressed. Potential use cases outlined by Tavus include tutoring, practicing difficult conversations, and camera-based technical support.
Founded in 2020 in San Francisco, Tavus has raised approximately $64 million in funding. The startup initially focused on building personalized artificial intelligence videos for sales and marketing before expanding into real-time interactive video conversations with digital personas. This breakthrough in generative video and conversational latency highlights the rapid advancement of multimodal communication models.