Tavus Unveils Griffin: 48% Mistook the AI Interviewer for Human

What happened

Tavus announced Griffin on October 10. According to the company, Griffin generates every pixel of every frame in real time from a single reference image — not just the face, arms and fingers, but also the chair's movement, the shadows it casts and the background. The avatars come in male and female looks and breathe and blink naturally. In a 54-person mock job-interview test, 48% believed the interviewer was human; among those confident in their judgment, the mistaken group averaged 79% confidence and the correct group 81%.

Key facts

1. 48% fooled, 54 participants.Per Tavus, nearly half mistook the AI interviewer for a human; the company says this shows the avatar passes the Turing Test.
2. Single-image real-time full-body generation.Every pixel of every frame rendered live from one reference image, covering expressions, gestures, chair movement and shadows; it can interrupt and be interrupted without losing its train of thought.
3. CEO statement.Tavus co-founder and CEO Hassaan Raza said in the announcement that humans are evolutionarily designed for face-to-face communication, where expressions, tone, gestures and timing matter as much as words; the new model removes past AI avatar giveaways, such as smiling while hearing bad news.
4. Demos and uses.In demos the model teaches Rubik's Cube solving and plays Simon Says, responding to visual cues; the company says it can be used for tutoring in schools.