Griffin was the first human interaction model that convinced me
Griffin realistically imitates voice, behavior, and movements, but an open version could make video call scams easier.
Griffin is an artificial intelligence model created by Tavus to understand and generate real-time video interactions. It follows speech, images, expressions, and pauses, and produces speech and video at the same time. The company describes Griffin-Lite as a research preview of the model.
For people talking with artificial intelligence, this could make the exchange feel more like an in-person conversation. Griffin can react while the other person is still speaking, interrupt or wait, change its tone and gestures to fit the context, and notice what appears on camera. Tavus reports that, in a live study, 48% of participants thought they had spoken with a real person.
Griffin-Lite is available to a select group of early testers; Tavus says it plans to release a more advanced model to a wider audience later. The source also does not explain the possibility of scams during video calls. It is therefore worth watching how access and usage limits are set before trusting the identity of someone seen on a call.
The link opens the original content; some require a subscription.
Go further on INEMA.CLUB
More news
- 15/0912 new AIs expected in October 2026What is rumor and what is already a verified release, in one guide.
- 02/09Humanoid robots getting more realistic: the UBTECH caseMore companies are showing human-like robots, and one of them fools you at first glance.
- 01/09Meta beats OpenAI and Google at real-time transcriptionWhat this changes for captions, meetings and voice support.
- 05/10More content about Claude Code MODsA page brings together content about Claude Code MODs.
- 05/10Files record Claude and Codex errors and limitsThe system reviews the records weekly and proposes lessons for approval before adding them to aprendizado.md.