Skip to content
worth noting New models

Company Tavus introduces the Griffin model for live video calls and opens limited Griffin-Lite preview

only one source so far

The company Tavus has introduced the Griffin model for live audiovisual conversations. According to its study, 48% of participants considered it human after a one-minute call. The Griffin-Lite model is available to selected testers; a more powerful version is to follow after safety questions have been resolved.

The company Tavus has introduced the Griffin model, which, according to the company, processes speech, facial expressions, tone of voice, gestures, and pauses during a conversation in real time. At the same time, it receives and generates video. The company describes it as the first "Human Interaction Model," i.e. a category of models intended for direct audiovisual conversations.

According to a Tavus study, 48% of participants considered the Griffin model to be human after a one-minute video call; previous systems achieved at most 2%. In a test that Tavus describes as an independent test by the company Nvidia focused on the humanness of audiovisual conversation, the Griffin model scored 3.83 points. Humans scored 3.92 points and the previous best AI model scored 2.80 points.

The Griffin-Lite model is available to selected testers as a research preview. According to the company, a more powerful version is to come after safety questions have been resolved. Tavus states possible uses as tutoring, practicing difficult conversations, and technical support using a camera.

What changed

Why it matters

Processing both voice and visual expressions could, according to the intended uses, expand the possibilities of practicing conversations and technical support via camera. The result of the Tavus study also shows the possibility of mistaking AI for a human during a short call. Practical use is for now limited by access restricted to selected testers only and by unresolved safety questions.

Release card

Griffin-Lite

Tavus

in the preview
Availability
The Griffin-Lite model is available to selected testers as a research preview.

The card summarizes information from the article and any dated corrections, with a link to the original source. It is not our assessment of the model. It does not yet have a dedicated editorial profile. Model selection and other announcements →

Relevant practical impact

What this means

01

For individuals

For individuals, the Griffin model is a possible future aid for tutoring or practicing difficult conversations. These possibilities are stated by the company Tavus; access to the Griffin-Lite model is currently limited to selected testers only.

More practical updates →
Griffin Griffin-Lite Tavus

Check the original

Event sources

only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.

1
The Decoder (daily AI news) independent context · first detected Nearly half of test subjects mistook Tavus' AI video avatar for a real person on a one-minute call