What if an AI could walk into a job interview or viva in your place and answer questions in real time? Tavus, a San Francisco-based company, has made this possible with its latest Human Interaction Model (HIM) model, Griffin. The AI tool is designed to have face-to-face video conversations using natural speech, facial expressions, gestures, and body movements. Moreover, unlike the regular chatbots, the new AI tool can understand both what people say and visual signals during a conversation and respond based on that. Follow along with the article to learn everything we know so far about Griffin.
Griffin is an AI model designed to take part in real-time video conversations by a San Francisco-based company known as Tavus. The company calls it the first Human Interaction Model (HIM). Unlike a chatbot that mainly responds to typed questions, Griffin is built for face-to-face interaction. It can listen to speech, watch visual cues and respond with a human-like voice, facial expressions and movements.
Not only that, but according to a claim made by the makers, the AI model scored 3.83 out of 5 on NVIDIA’s VideoFDB benchmark for video generation. Whereas the human reference scored 3.92, while the next-highest published system scored 2.80.
Introducing Griffin, the first model to pass the video Turing test.
— Tavus (@tavus) October 1, 2026
48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video.
It’s the first Human Interaction Model (HIM). pic.twitter.com/eb11XbC8Xr
The company also recently shared a post on X introducing its all-new Human Interaction Model (HIM), along with a video demonstrating the AI model in action. In the video, a person can be seen interacting with HIM, which responds to their queries easily, with no visible glitches or interruptions.
Moreover, the responses are accompanied by natural body movements, hand gestures, and facial expressions, making it increasingly difficult to tell whether you’re interacting with an AI model or a real human. Keep in mind that we’ve not officially tested the AI tool yet.
According to the company, Griffin continuously processes what is happening during a conversation, allowing it to react to interruptions, changes in tone, and other visible or audible signals while the discussion is still taking place.
Furthermore, the model builds on Tavus’ earlier work in video generation, speech, and conversational understanding. These systems help Griffin interpret facial expressions, tone, timing, interruptions, and the surrounding visual context.
The company believes these features could make Griffin useful for applications such as practice interviews, tutoring, and customer support, as users can interact with it naturally through conversation rather than having to learn specific commands or prompts.
Also read: OpenAI alerts over 100 organisations about rogue AI agent activity, check all details
At the time of writing, Tavus has not yet made Griffin openly available to everyone. The company said that they are working on disclosure and safety measures before making the technology widely available. However, it’s important to note that the company is currently offering Griffin-Lite to eligible testers for initial testing.