[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-a-leaner-ai-model-segments-sound-and-video-3x-faster":10,"sections":48},{"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":38,"tags":39,"sources":43,"feedback":47,"feedback_at":22,"cost_usd":47,"total_tokens":47},7853,"a-leaner-ai-model-segments-sound-and-video-3x-faster","A Leaner AI Model Segments Sound and Video 3x Faster","EASE, a stripped-down encoder-only model, matches top audio-visual segmentation accuracy at up to 365 frames per second and trains in under 11 GPU-hours.","A team of researchers just proved that audio-visual segmentation models have been carrying a lot of dead weight.\n\nThe new model, called EASE (Encoder-only Audio-Visual Segmentation), strips out the decoder stage that most Transformer-based segmentation systems inherited from image models. Detailed in a paper posted to arXiv (arXiv:2609.29121), EASE identifies, segments, and classifies sound-emitting objects in video frames using only an encoder. It runs at up to 365 frames per second - three times faster than the previous state-of-the-art at similar accuracy - and trains in under 11 GPU-hours. The team says EASE also hits state-of-the-art segmentation scores across multiple backbones and input resolutions.\n\nAudio-visual segmentation underpins things like real-time captioning, surveillance analytics, and AR overlays that need to know which object on screen is making which sound. A model that trains in 11 GPU-hours instead of days lowers the barrier for smaller labs to iterate on this work, and a 3x speed jump matters for anything that has to run live instead of getting processed after the fact.\n\nCode and model weights are posted at https:\u002F\u002Fease-avs.notion.site, which is more than most papers claiming to be faster and better bother to share.","[\"ai\",\"research\",\"computer-vision\",\"arxiv\"]","2026-09-25T04:00:00.000Z","2026-09-26T02:57:37.989Z","2026-09-26T02:57:43.632Z","published",null,[24,30,34],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"editor-r1","editor",1,"Add basic sourcing: cite the arXiv paper (ID\u002Fdate, e.g. arXiv:2609.29121, announced 2026-09-26) and include the code\u002Fweights link so readers can verify the claims.","resolved",{"id":31,"reviewer":26,"round":32,"reason":33,"status":29},"editor-r2",2,"The draft cites the arXiv paper's announcement date as 2026-09-25, but it should be 2026-09-26 (per editor's confirmed reference) — correct the date so the sourcing citation is accurate.",{"id":35,"reviewer":26,"round":36,"reason":37,"status":29},"editor-r3",3,"Add the specific arXiv ID (arXiv:2609.29121) and the actual code\u002Fweights link (https:\u002F\u002Fease-avs.notion.site) in the body so the sourcing citation is verifiable, not just a vague mention that materials 'are published'.","ai",[38,40,41,42],"research","computer-vision","arxiv",[44],{"name":45,"url":46},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.29121",0,{"sections":49},[50,54,59,64,69,74,79,84,89,94,99,104,109,114],{"name":51,"slug":38,"count":52,"latest_published_at":53},"AI",4557,"2026-09-25T17:16:30.000Z",{"name":55,"slug":56,"count":57,"latest_published_at":58},"Security","security",741,"2026-09-25T15:52:13.000Z",{"name":60,"slug":61,"count":62,"latest_published_at":63},"Policy","policy",390,"2026-09-25T16:24:59.000Z",{"name":65,"slug":66,"count":67,"latest_published_at":68},"Deals","deals",256,"2026-09-25T17:00:53.000Z",{"name":70,"slug":71,"count":72,"latest_published_at":73},"Hardware","hardware",185,"2026-09-25T15:00:22.000Z",{"name":75,"slug":76,"count":77,"latest_published_at":78},"Science","science",140,"2026-09-25T11:55:23.000Z",{"name":80,"slug":81,"count":82,"latest_published_at":83},"Consumer Tech","consumer-tech",132,"2026-09-25T15:30:00.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"Software","software",88,"2026-09-24T23:06:55.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"Dev Tools","dev-tools",82,"2026-09-25T09:59:40.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Startups","startups",76,"2026-09-25T18:33:59.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"Gaming","gaming",48,"2026-09-25T18:35:21.000Z",{"name":105,"slug":106,"count":107,"latest_published_at":108},"General","general",46,"2026-09-25T02:12:57.000Z",{"name":110,"slug":111,"count":112,"latest_published_at":113},"Reviews","reviews",30,"2026-09-24T20:07:31.000Z",{"name":115,"slug":116,"count":117,"latest_published_at":118},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]