{"rewrite":{"id":"r_f5a0d103001daaaa2585cb82","clusterId":"c_79d4d0461a3f1c75d11b6311","slug":"tavus-griffin-ai-fools-48-of-video-callers-into-thinking-it-is-human","model":"deepseek-v4-1-flash","headline":"Tavus Griffin AI Fools 48% of Video Callers Into Thinking It Is Human","summary":"Tavus announced Griffin, an AI model that handles real-time video and audio conversation. In a research preview test called Griffin-Lite, 26 of 54 participants on a one-minute live video call judged the AI to be a real human, about 48%. Griffin generates 720p video from a single reference image, including arms, hands, chair, shadows, and background, with an average delay of 0.43 seconds on NVIDIA H100 hardware.","whyItMatters":"A model that generates a full video presence from one still image, at sub-second latency, moves synthetic video past the avatar stage and into something a live call can carry without the other side noticing.","webCardHtml":"\u003cp\u003eTavus calls Griffin a Human Interaction Model. It reads facial expressions, gaze, tone, and silence at intervals under one second, then generates its own voice, expression, and gestures in parallel rather than waiting for the user to finish speaking. The company splits the system into a Continuous Conversational Modeling engine, which decides what to express and when, and an Audio-Visual Generation engine, which produces the audio and video. Tavus says NVIDIA ran the evaluation independently in September 2026.\u003c/p\u003e","blueskyPost":"Griffin's 0.43-second delay on H100 hardware is the number that decides whether Tavus can sell this to call centers. The 48% human-judgment figure matters less than whether that latency holds on cheaper chips.","twitterPost":"Griffin's 0.43-second delay on H100 hardware decides if Tavus can sell this to call centers, not the 48% figure.","threadsPost":"Tavus's 48% figure comes from a 54-person test. The number that decides Griffin's commercial future is the 0.43-second latency on NVIDIA H100 hardware, because that is the compute cost any buyer has to absorb at scale.","newsletterBlurb":"Tavus announced Griffin, an AI model that runs two-way video and audio conversation in real time. In a research preview test, 26 of 54 participants judged the AI to be human, about 48%. Griffin generates 720p video from a single reference image, and Tavus says NVIDIA ran the benchmark evaluation independently in September 2026.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20261006-tavus-griffin/\",\"title\":\"New AI model 'Griffin' makes 48% of video call participants mistake it for a human, recreating facial and body movements in video from a single image\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4664,"outputTokens":593,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1791295636,"createdAt":"2026-10-06T14:03:47.000Z","publishedAt":"2026-10-06T14:06:08.000Z","updatedAt":"2026-10-06T14:06:08.000Z"},"cluster":{"id":"c_79d4d0461a3f1c75d11b6311","canonicalTitle":"ビデオ通話で参加者の48％が人間だと誤認したAIモデル「Griffin」が登場、1枚の画像から顔や体の動きを映像で再現","representativeArticleId":"a_0b706186141a10a7516068bf","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[\"Griffin\"],\"studios\":[\"Tavus\"],\"people\":[],\"type\":\"news\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-10-06T13:00:00.000Z","lastSeenAt":"2026-10-06T13:00:00.000Z","updatedAt":"2026-10-06T14:06:06.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20261006-tavus-griffin/","title":"ビデオ通話で参加者の48％が人間だと誤認したAIモデル「Griffin」が登場、1枚の画像から顔や体の動きを映像で再現"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":["Griffin"],"studios":["Tavus"],"people":[],"type":"news","domain":"other","is_roundup":false},"keyFacts":null}