[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-deepfake-detectors-can-now-skip-most-of-the-video":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},8988,"deepfake-detectors-can-now-skip-most-of-the-video","Deepfake Detectors Can Now Skip Most of the Video","A lightweight router previews video and audio clips so deepfake detectors only analyze the few seconds likely to be faked, cutting compute nearly 16x.","A new detector doesn't bother watching most of the video, and gets better results for it.\n\nResearchers built ModalFidelity, a lightweight router that previews each one-second window of a video's audio and image streams before any forensic detector runs. It flags only the windows worth deeper scrutiny, under a compute budget it can never exceed. Tested on the AV-Deepfake1M benchmark, the router examined at most one-fifth of all windows yet beat systems that filter after running full detection, using 15.9 times less compute. It kept more than 96% of the accuracy of an oracle that already knows exactly where every forgery sits.\n\nThe trick matters because deepfakes have gotten smaller, not bigger. Generators now read a transcript and alter only the handful of seconds where a sentence's meaning flips, leaving the rest of the clip untouched. Detectors that still scan every second of audio and video are wasting almost all their compute on footage that was never faked, a mismatch ModalFidelity is built to exploit.\n\nIt's a clever reframing: deciding where to look is cheap, looking is expensive, so do the cheap part first. Whether that logic holds once forgers start optimizing against detectors that are known to skim is the open question nobody's benchmark can answer yet.","[\"deepfake-detection\",\"ai-research\",\"computer-vision\",\"compute-efficiency\"]","2026-10-01T04:00:00.000Z","2026-10-01T14:30:33.007Z","2026-10-01T14:30:36.882Z","published",null,[],"ai",[26,27,28,29],"deepfake-detection","ai-research","computer-vision","compute-efficiency",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.38246",0,{"sections":36},[37,40,44,49,54,59,63,68,73,77,82,87,92,97],{"name":38,"slug":24,"count":39,"latest_published_at":18},"AI",5488,{"name":41,"slug":42,"count":43,"latest_published_at":18},"Security","security",809,{"name":45,"slug":46,"count":47,"latest_published_at":48},"Policy","policy",429,"2026-10-01T02:26:17.000Z",{"name":50,"slug":51,"count":52,"latest_published_at":53},"Deals","deals",298,"2026-09-30T21:00:26.000Z",{"name":55,"slug":56,"count":57,"latest_published_at":58},"Hardware","hardware",196,"2026-09-30T13:00:00.000Z",{"name":60,"slug":61,"count":62,"latest_published_at":18},"Science","science",162,{"name":64,"slug":65,"count":66,"latest_published_at":67},"Consumer Tech","consumer-tech",149,"2026-09-30T22:57:11.000Z",{"name":69,"slug":70,"count":71,"latest_published_at":72},"Dev Tools","dev-tools",93,"2026-10-01T02:30:48.000Z",{"name":74,"slug":75,"count":71,"latest_published_at":76},"Software","software","2026-09-30T21:41:11.000Z",{"name":78,"slug":79,"count":80,"latest_published_at":81},"Startups","startups",84,"2026-09-30T20:39:09.000Z",{"name":83,"slug":84,"count":85,"latest_published_at":86},"Gaming","gaming",51,"2026-09-30T16:24:30.000Z",{"name":88,"slug":89,"count":90,"latest_published_at":91},"General","general",50,"2026-09-30T21:37:54.000Z",{"name":93,"slug":94,"count":95,"latest_published_at":96},"Reviews","reviews",31,"2026-09-28T14:31:34.000Z",{"name":98,"slug":99,"count":100,"latest_published_at":101},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]