Nearly half of test subjects mistook Tavus' AI avatar for a real person on a one-minute call

Tavus, a San Francisco startup, has introduced Griffin, which it calls the first "Human Interaction Model" (HIM). The company says the model is built to understand and carry on face-to-face conversations in real time: it processes speech, facial expressions, tone of voice, gestures and pauses while both receiving and generating video.

48 percent took the avatar for a real person

In a Tavus study, 48 percent of participants believed Griffin was a real person after a one-minute video call. According to the company, previous systems never scored above two percent on that measure.

In what Tavus describes as an independent Nvidia test of how human an AI feels in a direct audio-video conversation, Griffin scored 3.83 points. Actual humans scored 3.92, while the previous best AI model reached 2.80.

A research preview for select testers

A preview version called Griffin-Lite is available to "select testers as a research preview." A more capable version will follow once "safety concerns are addressed," Tavus says.

The company lists tutoring, practicing difficult conversations and camera-based tech support as potential use cases for the model.

From sales videos to live personas

Tavus was founded in 2020 and has raised about $64 million. It started out building personalized AI videos for sales and marketing, then expanded into live video conversations with digital personas. More details on Griffin are in the Tavus research report.