Griffin is a video model from Tavus that holds live, face-to-face conversations as a human-like person on screen. Tavus calls it a Human Interaction Model.
In plain words
Griffin joins seeing and hearing you, holding the conversation and producing a lifelike face in one system that takes video in and sends video out. It listens while you speak and notices your expressions and pauses as well as your words. In a live Tavus study, 48% of people who talked with it believed they had spoken to a real person. For now, only a small group of early testers can use a research preview called Griffin-Lite.
A quick example
Imagine a language school that wants students to practise speaking with a tutor who appears on a video call and reacts when they hesitate. A model like Griffin could play that tutor, while a teacher reviews the recorded sessions.
Where you'll meet it
- Video agents that talk with people face to face in real time.
- Research on how well AI handles nonverbal signals such as pauses and expressions.
Related
Sources
- DocsGriffin: The First Human Interaction Model, Tavus · read 4 Oct 2026
- PaperVideoFDB: Evaluating Full-Duplex Vision-Speech Capabilities in Conversational Agents, NVIDIA Research · read 4 Oct 2026
Checked by WhatIsAI against the sources above · 4 Oct 2026. A full five-minute explainer with a diagram may follow. Spotted an error? Report it.