In this work, we present LiveAnimate, which brings 14B-parameter video Diffusion Transformers to real-time streaming inference (19.63 FPS on 2x H100) for interactive full-body animation. We focus on solving the quality degradation and memory explosion typical of long-form diffusion generation. We do this through two key designs: a Reference-Anchored training pipeline with block-wise distillation to hit a 3-step sampling budget, and PR-Sink Attention, a bounded KV-cache mechanism that keeps latency and memory constant over long streams by retrieving historical appearance context based on pose similarity.</p>\n","updatedAt":"2026-08-14T05:11:30.970Z","author":{"_id":"64489bb5e21484883408a96d","avatarUrl":"/avatars/23f7aa4d733a7708fab4ee059ac2b323.svg","fullname":"Jinpeng YU","name":"Jacob-Yu","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":3,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8900464773178101},"editors":["Jacob-Yu"],"editorAvatarUrls":["/avatars/23f7aa4d733a7708fab4ee059ac2b323.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.11745","authors":[{"_id":"6a7ea06a42823931a1f1772f","name":"Yuxuan Zhang","hidden":false},{"_id":"6a7ea06a42823931a1f17730","name":"Haozhong Xiong","hidden":false},{"_id":"6a7ea06a42823931a1f17731","name":"Yubo Huang","hidden":false},{"_id":"6a7ea06a42823931a1f17732","name":"Jiayi Song","hidden":false},{"_id":"6a7ea06a42823931a1f17733","user":{"_id":"64489bb5e21484883408a96d","avatarUrl":"/avatars/23f7aa4d733a7708fab4ee059ac2b323.svg","isPro":false,"fullname":"Jinpeng YU","user":"Jacob-Yu","type":"user","name":"Jacob-Yu"},"name":"Jinpeng Yu","status":"claimed_verified","statusLastChangedAt":"2026-08-14T08:45:04.765Z","hidden":false},{"_id":"6a7ea06a42823931a1f17734","name":"Haofan Wang","hidden":false},{"_id":"6a7ea06a42823931a1f17735","name":"Jiaming Liu","hidden":false},{"_id":"6a7ea06a42823931a1f17736","name":"Ruihua Huang","hidden":false},{"_id":"6a7ea06a42823931a1f17737","name":"Liwei Wang","hidden":false}],"publishedAt":"2026-08-13T00:00:00.000Z","submittedOnDailyAt":"2026-08-14T00:00:00.000Z","title":"LiveAnimate: Stable Long-Form Streaming Human Animation in Real-Time","submittedOnDailyBy":{"_id":"64489bb5e21484883408a96d","avatarUrl":"/avatars/23f7aa4d733a7708fab4ee059ac2b323.svg","isPro":false,"fullname":"Jinpeng YU","user":"Jacob-Yu","type":"user","name":"Jacob-Yu"},"summary":"Pose-driven human animation synthesizes a video of a target person from a single reference image and a driving pose stream. Real-time generation is essential for interactive applications such as live streaming, telepresence, and virtual avatars, yet diffusion-based systems require minutes to hours per clip, precluding responsive interaction. We present LiveAnimate, to our knowledge the first animation system to combine real-time streaming with stable long-form generation at billion scale, built on a 14B-parameter video Diffusion Transformer (DiT). A two-stage training pipeline first adapts a pretrained bidirectional DiT into a block-causal autoregressive generator through Reference-Anchored Teacher-Forcing Adaptation, and then reduces the sampling budget to three steps through Block-wise Self-Forcing Distillation. To preserve appearance over extended streams, we introduce Pose-Retrieval Sink Attention (PR-Sink), a bounded KV-cache mechanism combining a Static Sink that permanently anchors the first generated block, a Dynamic Sink that holds a pose-retrieved historical block, and a three-slot Rolling Window. When a pose recurs, PR-Sink restores the relevant appearance context without retaining the entire sequence, so memory and per-block latency remain constant regardless of stream duration. Together with Ulysses sequence parallelism and operator fusion, these designs enable 19.63\\,FPS streaming inference on two NVIDIA H100 GPUs. On a three-minute benchmark, LiveAnimate maintains nearly constant perceptual quality and identity from the first 30 seconds to the final minute, while prior systems degrade substantially or require hours of offline computation for the same rollout. These results establish a new operating point in quality, latency, and duration for interactive full-body animation.","upvotes":9,"discussionId":"6a7ea06a42823931a1f17738","projectPage":"https://liveanimate.github.io/","githubRepo":"https://github.com/liveanimate/LiveAnimate","githubRepoAddedBy":"user","ai_summary":"LiveAnimate enables real-time, long-form pose-driven human animation via a 14B-parameter video diffusion transformer with specialized training, bounded attention caching, and sequence parallelism.","ai_keywords":["Diffusion Transformer (DiT)","Reference-Anchored Teacher-Forcing Adaptation","Block-wise Self-Forcing Distillation","Pose-Retrieval Sink Attention (PR-Sink)","Static Sink","Dynamic Sink","Rolling Window","Ulysses sequence parallelism","operator fusion"],"ai_summary_model":"thinkingmachines/Inkling-Small","githubStars":7,"organization":{"_id":"6a6841e7107886ba1a151b03","name":"QwenBusinessUnit","fullname":"Qwen Business Unit","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/66f79b323fe089b75e9e0c04/MlefZsdry-JuhKzAwxjQl.webp"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"64489bb5e21484883408a96d","avatarUrl":"/avatars/23f7aa4d733a7708fab4ee059ac2b323.svg","isPro":false,"fullname":"Jinpeng YU","user":"Jacob-Yu","type":"user"},{"_id":"66d53350ad293ffc4b178d10","avatarUrl":"/avatars/d1cb7c800d2af4081de62ef95dc2c44c.svg","isPro":false,"fullname":"jackylova","user":"jackylova","type":"user"},{"_id":"688cceac29c69ec01eb28644","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/wyL1U86pB5GIi4x3JflOa.png","isPro":false,"fullname":"Chuyue Li","user":"woody-woody","type":"user"},{"_id":"67051ad602b48c3fce373fe7","avatarUrl":"/avatars/ebf10382f7c34f41b77f30eb1eea784e.svg","isPro":false,"fullname":"sjy","user":"sjy92","type":"user"},{"_id":"636b3f9ce3ad78bc68b67541","avatarUrl":"/avatars/2b7e745953ae39e01222e99fb63b279e.svg","isPro":false,"fullname":"yuxuan","user":"zzyx","type":"user"},{"_id":"68f9a5cf4cbb5261e6a3cf78","avatarUrl":"/avatars/a0e9bcadd2b69e35ccca6f7d75c2b0b0.svg","isPro":false,"fullname":"Ethan Miller","user":"YaKaYC","type":"user"},{"_id":"69be48e4df85f5f792fc2a7e","avatarUrl":"/avatars/06505fd14a44973c6305ec3f23a67ee4.svg","isPro":false,"fullname":"None","user":"Gonzalo320","type":"user"},{"_id":"65c4eb7cd1dcbd30d86febec","avatarUrl":"/avatars/001c8f02e8ce794b2c21883628b2da72.svg","isPro":false,"fullname":"free-bit","user":"free-bit","type":"user"},{"_id":"620783f24e28382272337ba4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620783f24e28382272337ba4/zkUveQPNiDfYjgGhuFErj.jpeg","isPro":false,"fullname":"GuoLiangTang","user":"Tommy930","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"6a6841e7107886ba1a151b03","name":"QwenBusinessUnit","fullname":"Qwen Business Unit","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/66f79b323fe089b75e9e0c04/MlefZsdry-JuhKzAwxjQl.webp"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.11745.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2608.11745 in a model README.md to link it from this page.

Cite arxiv.org/abs/2608.11745 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2608.11745 in a Space README.md to link it from this page.

Add this paper to a collection to link it from this page.