Griffin generates video at a resolution of 720p in 350 ms chunks, according to Tavus
According to Tavus, Griffin generates video at a resolution of 720p in 350 ms chunks and responds to speech and visual input during a call. Griffin-Lite is available to selected testers as a research preview.
Newly published specifications for Griffin describe continuous video generation at a resolution of 720p in 350 ms chunks. According to Tavus, the system responds to speech and visual input in real time during a video call and accompanies its communication with gestures.
Tavus introduced Griffin for live audiovisual conversations. The model simultaneously receives and generates video and processes speech, facial expressions, tone of voice, gestures and pauses. In a study by Tavus, 48 % of participants thought Griffin was human after a one-minute video call; this is a result published by the company.
Griffin-Lite is available to selected testers as a research preview. According to the company, a more capable version is due to arrive once safety issues have been resolved. Tavus lists tutoring, practising difficult conversations and technical support via a camera as possible uses.
Why it matters
Responses to visual input may be useful for tasks where speech alone is not enough, such as showing a technical problem on camera. Tavus presents this use as a possibility; availability to selected testers currently allows evaluation within a limited research preview.
Release card
Griffin-Lite
Tavus
- Availability
- Griffin-Lite 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 →
What was added since the original report
Verified updates
-
The system generates video at a resolution of 720p.; Video is generated in 350 ms chunks.
- The system generates video at a resolution of 720p.
- Video is generated in 350 ms chunks.
Two audiences, two different impacts
What this means
For individuals
The result of a study by Tavus suggests that natural speech and gestures during a short video call may not reliably distinguish a human from AI.
More practical updates →For a business
For teams considering technical support using a camera, this currently offers an opportunity for limited testing. The company makes availability of the more capable version conditional on resolving safety issues, which limits planning for wider deployment.
ProcessesCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.