EverAnimate code, 480p LoRA checkpoints, minimal demo data, and training/inference scripts are now released: <a href=\"https://huggingface.co/epfl-vita/everanimate\">https://huggingface.co/epfl-vita/everanimate</a></p>\n","updatedAt":"2026-05-27T17:32:02.544Z","author":{"_id":"62b705a6dd998a8b1e422936","avatarUrl":"/avatars/ab307b6548c8c627d640635f3316c5ad.svg","fullname":"wuyang li","name":"wymanCV","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7719546556472778},"editors":["wymanCV"],"editorAvatarUrls":["/avatars/ab307b6548c8c627d640635f3316c5ad.svg"],"reactions":[],"isReport":false}},{"id":"6a172b76141b559c8e5db8c9","author":{"_id":"62b705a6dd998a8b1e422936","avatarUrl":"/avatars/ab307b6548c8c627d640635f3316c5ad.svg","fullname":"wuyang li","name":"wymanCV","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false},"createdAt":"2026-05-27T17:35:50.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"Resources:\n- Code: https://github.com/vita-epfl/EverAnimate\n- Project page: https://everanimate.github.io/homepage/\n- Model and demo data: https://huggingface.co/epfl-vita/everanimate","html":"<p>Resources:</p>\n<ul>\n<li>Code: <a href=\"https://github.com/vita-epfl/EverAnimate\" rel=\"nofollow\">https://github.com/vita-epfl/EverAnimate</a></li>\n<li>Project page: <a href=\"https://everanimate.github.io/homepage/\" rel=\"nofollow\">https://everanimate.github.io/homepage/</a></li>\n<li>Model and demo data: <a href=\"https://huggingface.co/epfl-vita/everanimate\">https://huggingface.co/epfl-vita/everanimate</a></li>\n</ul>\n","updatedAt":"2026-05-27T17:35:50.558Z","author":{"_id":"62b705a6dd998a8b1e422936","avatarUrl":"/avatars/ab307b6548c8c627d640635f3316c5ad.svg","fullname":"wuyang li","name":"wymanCV","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.630043625831604},"editors":["wymanCV"],"editorAvatarUrls":["/avatars/ab307b6548c8c627d640635f3316c5ad.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2605.15042","authors":[{"_id":"6a170edbda9422d403a421e3","name":"Wuyang Li","hidden":false},{"_id":"6a170edbda9422d403a421e4","name":"Yang Gao","hidden":false},{"_id":"6a170edbda9422d403a421e5","name":"Mariam Hassan","hidden":false},{"_id":"6a170edbda9422d403a421e6","name":"Lan Feng","hidden":false},{"_id":"6a170edbda9422d403a421e7","name":"Wentao Pan","hidden":false},{"_id":"6a170edbda9422d403a421e8","name":"Po-Chien Luan","hidden":false},{"_id":"6a170edbda9422d403a421e9","name":"Alexandre Alahi","hidden":false}],"publishedAt":"2026-05-14T00:00:00.000Z","submittedOnDailyAt":"2026-05-27T00:00:00.000Z","title":"EverAnimate: Minute-Scale Human Animation via Latent Flow Restoration","submittedOnDailyBy":{"_id":"62b705a6dd998a8b1e422936","avatarUrl":"/avatars/ab307b6548c8c627d640635f3316c5ad.svg","isPro":false,"fullname":"wuyang li","user":"wymanCV","type":"user","name":"wymanCV"},"summary":"We propose EverAnimate, an efficient post-training method for long-horizon animated video generation that preserves visual quality and character identity. Long-form animation remains challenging because highly dynamic human motion must be synthesized against relatively static environments, making chunk-based generation prone to accumulated drift: (i) low-level quality drift, such as progressive degradation of static backgrounds, and (ii) high-level semantic drift, such as inconsistent character identity and view-dependent attributes. To address this issue, EverAnimate restores drifted flow trajectories by anchoring generation to a persistent latent context memory, consisting of two complementary mechanisms. (i) Persistent Latent Propagation maintains a context memory across chunks to propagate identity and motion in latent space while mitigating temporal forgetting. (ii) Restorative Flow Matching introduces an implicit restoration objective during sampling through velocity adjustment, improving within-chunk fidelity. With only lightweight LoRA tuning, EverAnimate outperforms state-of-the-art long-animation methods in both short- and long-horizon settings: at 10 seconds, it improves PSNR/SSIM by 8%/7% and reduces LPIPS/FID by 22%/11%; at 90 seconds, the gains increase to 15%/15% and 32%/27%, respectively.","upvotes":1,"discussionId":"6a170edbda9422d403a421ea","ai_summary":"EverAnimate addresses long-horizon animated video generation challenges through persistent latent propagation and restorative flow matching to maintain visual quality and character identity.","ai_keywords":["post-training method","long-horizon animated video generation","visual quality","character identity","chunk-based generation","accumulated drift","latent context memory","persistent latent propagation","restorative flow matching","velocity adjustment","LoRA tuning","PSNR","SSIM","LPIPS","FID"]},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"69bd46d749578ed0f29b5720","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/Ov44AQmjlmMOc4FIcgP9F.png","isPro":false,"fullname":"周晨曦","user":"liuus62","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2605/2605.15042.md"}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2605.15042 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2605.15042 in a Space README.md to link it from this page.

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