{"slug": "tavus-claims-48-pass-rate-if-video-turing-test", "title": "Tavus claims 48% pass rate if video Turing test", "summary": "Tavus announced Griffin, which it calls the first model to pass the video Turing test, with 48% of people who talked to it live believing it was a real human, versus a pass rate below 3% for previous systems. Tavus said Griffin ranks first on NVIDIA's benchmark for full-duplex AI video and is the first Human Interaction Model (HIM), generating entire frames rather than just a face while listening, watching and talking simultaneously. Tavus stated that because Griffin can be mistaken for a real person, it requires safety considerations before public release, detailed in a full research preview.", "body_md": "Tavus on X: \"Introducing Griffin, the first model to pass the video Turing test.\n48% 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.\nIt’s the first Human Interaction Model (HIM).\" / X\n\nTavus on X: \"Introducing Griffin, the first model to pass the video Turing test.\n48% 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.\nIt’s the first Human Interaction Model (HIM).\"\n\nIntroducing Griffin, the first model to pass the video Turing test.\n48% 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.\nIt’s the first Human Interaction Model (HIM).\n\nIntroducing Griffin, the first model to pass the video Turing test.\n48% 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.\nIt’s the first Human Interaction Model (HIM).\n\nWhen machines meet us where we are and truly understand us, we unlock a future where using a machine will be as easy as talking to a friend or coworker. A tutor for every student that notices when they’re lost, an elderly care companion that listens, or the perfect assistant.\n\nWatch Brittney and Mars try to trip Griffin up at Simon Says. To play, it has to listen for \"Simon says,\" watch their hands and move its own, all at the same time. It generates the whole frame, not just a face.\n\nWhile Mars is still telling Griffin about a boring date, it laughs, nods and asks questions instead of waiting for her to stop. That's full duplex: listening, watching and talking at the same time, the way people actually talk.\n\nMid-story, Mars holds up a peacock, and Griffin seamlessly writes it into the plot. It sees while it speaks, so what it sees can change what it's saying. Show it something and it's part of the conversation.\n\nAri solves a Rubik's cube while Griffin coaches him from what it sees in his hands, and when he goes quiet to think, it waits. Its perception is always on, so it acts without being asked. Note: this actually blew our minds seeing happen live.\n\nSee how Sagar keeps his eyes on the soldering iron while Griffin talks him through the job and counts the seconds for him. Time is part of the conversation too, not just words. That's what a good coach does without being asked.\n\nMost AI can tell you it's happy for you. Griffin shows it, in its voice, its face and its gestures, all at once. Rudy just got promoted, and you don't need the sound on to know how Griffin feels about it.\n\nBecause it’s the first model that can be mistaken for a real person, Griffin requires safety considerations before it can be released publicly. The full research preview covers how it works, how we tested it and the safety measures that have to come first:", "url": "https://wpnews.pro/news/tavus-claims-48-pass-rate-if-video-turing-test", "canonical_source": "https://twitter.com/tavus/status/2105704169009246248", "published_at": "2026-10-03 02:30:18+00:00", "updated_at": "2026-10-03 03:06:13.039282+00:00", "lang": "en", "topics": ["artificial-intelligence", "generative-ai", "ai-safety", "computer-vision", "ai-research"], "entities": ["Tavus", "Griffin", "NVIDIA", "Human Interaction Model", "Brittney", "Mars", "Ari", "Sagar"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/tavus-claims-48-pass-rate-if-video-turing-test", "markdown": "https://wpnews.pro/news/tavus-claims-48-pass-rate-if-video-turing-test.md", "text": "https://wpnews.pro/news/tavus-claims-48-pass-rate-if-video-turing-test.txt", "jsonld": "https://wpnews.pro/news/tavus-claims-48-pass-rate-if-video-turing-test.jsonld"}}