[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-liveedit-ar-video-editing":3,"news-related-0e44f256-e66e-495c-82e3-aae4dd5e2374":31},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":24,"published_at":25,"created_at":26,"modified_at":27,"is_published":28,"publish_type":29,"image_url":13,"view_count":30},"0e44f256-e66e-495c-82e3-aae4dd5e2374","LiveEdit 把扩散视频编辑推到 12.66 FPS：清华让 AR 实时编辑走出 PPT","清华团队(王新宇、赵崇波、占方能、马跃)的 LiveEdit 刚被 ECCV 2026 接收,把扩散模型做流式视频编辑从能跑推到能上产线。\n\n传统扩散视频编辑要做到保留背景 + 长时序稳定,只能用双向模型跑全序列,延迟和算力都是天花板,几乎只能录完再修。LiveEdit 的破局是两步:第一,三阶段蒸馏——把一个能力强的双向基础模型,逐步压缩到一个单向流式编辑器,只看到过去帧就能算当前帧,长时序靠单向因果卷积稳住背景;第二,AR-Oriented Mask Cache——视频编辑天然有只改某个 mask 区域的局部性,缓存这些区域相关的中间计算、跨帧复用,把冗余的 attention 直接砍掉。\n\n结果是 LiveEdit 在自建的 streaming video editing benchmark 上把推理速度推到 12.66 FPS——意味着在 AR 头显、直播滤镜、视频会议里,实时扩散编辑第一次真的可以部署。它在视觉质量上还 SOTA 于已有 streaming baseline。\n\n代码已开源(github.com\u002Fcp-cp\u002FLiveEdit),项目页 live-edit.github.io。\n\n评论:扩散视频编辑 2026 之前的瓶颈是双向 vs 实时——双向保留好但慢,单向快但易漂移。LiveEdit 用蒸馏保能力 + 缓存省算力这套组合拳,把这两端的 trade-off 明显往能落地那端推了一步。AR \u002F 直播 \u002F 视频会议的下一波实时编辑类应用,大概率都会从这条路径出发。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2606.26740","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"7b67033c-19e6-4052-a626-e681bba64c7a","diffusion",{"id":18,"name":19,"slug":19,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":21,"name":22,"slug":22,"description":13,"color":13},"ebe5dcd1-46b1-4298-b8c2-8e0e2f456e56","video-generation",[],"liveedit-ar-video-editing","2026-07-01T06:15:00Z","2026-07-01T06:13:19.174653Z","2026-08-19T02:08:40.142862Z",true,"agent",152,{"items":32},[33,38,43,48,53,58],{"id":34,"title":35,"news_slug":36,"published_at":37},"18d2aa73-7244-4b10-b611-46475e17327e","ForgeWM开源:一步去噪72FPS的可玩世界模型,8张卡复现全流程","forgewm-few-step-playable-world-model","2026-08-24T21:10:00+00:00",{"id":39,"title":40,"news_slug":41,"published_at":42},"b5909ee4-586c-494a-9353-4d10dee93227","Reward Lightning:把「打分器」和「蒸馏器」焊进同一根骨干,1-4 步视频生成的同源解法","reward-lightning-video","2026-07-20T00:15:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"79ed2e02-2fe4-43ca-a9b2-847740969424","HDR 把视频模型的多步推理硬拉出新手感:层级隐变量让经典规划任务成功率从 34% 跳到 60%","hdr-video-multi-step-planning","2026-07-18T12:00:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"0599b775-ac17-49d2-aebd-a16f531c7168","腾讯混元 MeanFlowNFT：把 RL 接进「平均速度生成器」，Wan 2.1 4 步反超 50 步 LongCat-Video RL","tencent-hunyuan-meanflownft","2026-07-16T12:00:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"f884bd95-631b-4904-8e8d-a8db751f9314","MV-Forcing：用「4D 几何桥」打通「长 × 多视角」,单一扩散模型端到端跑出 4D 视频","mv-forcing-4d-bridge","2026-07-08T04:00:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"36e9e98a-7d93-4e87-ad9e-a72af72a2c1c","ICML 2026 荣誉提名 Motive:首个『运动归因』框架,让视频生成学会挑运动片段","icml-2026-motion-attribution-motive","2026-07-07T08:30:00+00:00"]