← Back to Model Beat
Research·Oct 1·all news from October 1, 2026

Nearly half of test subjects mistook Tavus' AI video avatar for a real person on a one-minute call

Tavus has introduced a real-time AI video model called Griffin, which replicates human facial expressions, gestures, and tone during video calls. In a company study, 48 percent of participants could not distinguish the AI from a human during a one-minute interaction. This development highlights the increasing capability of synthetic video technology to mimic human presence, raising questions about the future of digital identity verification and interpersonal communication.

Covered by 1 source

Related stories

ResearchAI Solves a Major Unsolved Math Problem. Not Everyone Is HappyOct 4 · 4 sourcesResearchAI Whistleblowers, Google, OpenAI, Meta to Face New York City CouncilOct 4 · 12 sourcesResearchAI beats Stratego's greatest player, ending one of the last human strongholds in board gamesOct 1 · 3 sourcesResearchRLTL;DR: Self-Improvement by Internalizing Self-Generated FeedbackSep 29 · 4 sources