AI Griffin Passes Video Turing Test with 48% Score

Tavus has introduced the Griffin model, designed for realistic video communication using artificial intelligence.

In blind tests, 48% of participants mistakenly identified the avatar as a real person. For previous systems, this figure did not exceed 3%.

The technology generates not just the face, but the entire video frame in real time.

• Speech and visual context recognition of the interlocutor
• Tracking user actions during the dialogue
• Instant reaction to events in the frame