Hugging Face Daily Papers · July 3, 2026 · 3 min read

WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory

Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.

Like Read original ↗

Project Page: <a href=\"https://worlddirector.github.io/\" rel=\"nofollow\">https://worlddirector.github.io/</a><br>Github: <a href=\"https://github.com/pPetrichor/WorldDirector\" rel=\"nofollow\">https://github.com/pPetrichor/WorldDirector</a></p>\n","updatedAt":"2026-07-03T04:47:30.339Z","author":{"_id":"665f059a8947302aa2c63afe","avatarUrl":"/avatars/50f560285946532321a0bd526494148d.svg","fullname":"hanlin wang","name":"hlwang06","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":26,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.606141984462738},"editors":["hlwang06"],"editorAvatarUrls":["/avatars/50f560285946532321a0bd526494148d.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2607.02517","authors":[{"_id":"6a471b2f6ee372f6920de2b6","name":"Hanlin Wang","hidden":false},{"_id":"6a471b2f6ee372f6920de2b7","name":"Hao Ouyang","hidden":false},{"_id":"6a471b2f6ee372f6920de2b8","name":"Qiuyu Wang","hidden":false},{"_id":"6a471b2f6ee372f6920de2b9","name":"Wen Wang","hidden":false},{"_id":"6a471b2f6ee372f6920de2ba","name":"Qingyan Bai","hidden":false},{"_id":"6a471b2f6ee372f6920de2bb","name":"Ka Leong Cheng","hidden":false},{"_id":"6a471b2f6ee372f6920de2bc","name":"Yue Yu","hidden":false},{"_id":"6a471b2f6ee372f6920de2bd","name":"Yixuan Li","hidden":false},{"_id":"6a471b2f6ee372f6920de2be","name":"Yihao Meng","hidden":false},{"_id":"6a471b2f6ee372f6920de2bf","name":"Zichen Liu","hidden":false},{"_id":"6a471b2f6ee372f6920de2c0","name":"Yanhong Zeng","hidden":false},{"_id":"6a471b2f6ee372f6920de2c1","name":"Yujun Shen","hidden":false},{"_id":"6a471b2f6ee372f6920de2c2","name":"Qifeng Chen","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/6039478ab3ecf716b1a5fd4d/7-X_Hg3naThHm70xy-iLS.mp4"],"publishedAt":"2026-07-02T00:00:00.000Z","submittedOnDailyAt":"2026-07-03T00:00:00.000Z","title":"WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory","submittedOnDailyBy":{"_id":"6039478ab3ecf716b1a5fd4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg","isPro":true,"fullname":"taesiri","user":"taesiri","type":"user","name":"taesiri"},"summary":"We present WorldDirector, a highly controllable video world model framework designed for persistent dynamic object memory and unrestricted viewpoint exploration. Unlike existing world models that entangle physical dynamics with pixel rendering and rely on continuous visual observation to sustain motion, our framework explicitly decouples semantic motion orchestration from visual generation. By leveraging an LLM to coordinate 3D trajectories with camera movements and subsequently employing these orchestrated trajectories as control signals for video generation, our approach ensures strict physical logic and appearance stability, successfully preserving the exact visual identities of dynamic entities even when they re-enter the scene after prolonged periods out of view. Experimental results demonstrate that our method supports the synthesis of complex and extended events with unprecedented controllability and persistent dynamic object memory. Project Page: https://worlddirector.github.io/","upvotes":11,"discussionId":"6a471b2f6ee372f6920de2c3","ai_summary":"WorldDirector enables controllable video generation with persistent object memory by decoupling semantic motion planning from visual rendering through LLM coordination of 3D trajectories and camera movements.","ai_keywords":["video world model","persistent dynamic object memory","viewpoint exploration","semantic motion orchestration","visual generation","LLM","3D trajectories","camera movements","physical logic","appearance stability"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct"},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"665f059a8947302aa2c63afe","avatarUrl":"/avatars/50f560285946532321a0bd526494148d.svg","isPro":false,"fullname":"hanlin wang","user":"hlwang06","type":"user"},{"_id":"64acd2ec39fcfebff8c79c00","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/64acd2ec39fcfebff8c79c00/Avq66l5hO-aggNtk4Y1ss.png","isPro":false,"fullname":"Ka Leong Cheng","user":"felixcheng97","type":"user"},{"_id":"646c37261da1b6d027ff20ef","avatarUrl":"/avatars/2c2feb9b3f29169ade74632a95450175.svg","isPro":false,"fullname":"hao","user":"GPTVer8","type":"user"},{"_id":"6478a982256b62e219917d67","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/PUJ-N2cQxgEmDGfyjajyA.jpeg","isPro":false,"fullname":"JingyeChen22","user":"JingyeChen22","type":"user"},{"_id":"67da745e0f5863ac3197b801","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/4tpzA6yjkU3DA-XhydezI.png","isPro":false,"fullname":"zhou nan","user":"rescuerz","type":"user"},{"_id":"6a2da6c8ca070ee12c6e396c","avatarUrl":"/avatars/0355287dcabaa67dbc7f0b10b87451f9.svg","isPro":false,"fullname":"Joe Mama","user":"JoeMama123123123","type":"user"},{"_id":"660395bd738bdf26e74f120d","avatarUrl":"/avatars/65a09aa15ebf612492f132b15f20c592.svg","isPro":false,"fullname":"WANG Yifan","user":"ywangpa","type":"user"},{"_id":"655c2cc7e8a8971e89aa6e75","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/655c2cc7e8a8971e89aa6e75/rWQEbTqgq7MLEJJGfjP5y.jpeg","isPro":false,"fullname":"LIU Zichen","user":"LiuZichen","type":"user"},{"_id":"643815c4961bb61e463c5896","avatarUrl":"/avatars/3b44592472f16c56105bff8c314d9939.svg","isPro":false,"fullname":"Jianxiong Gao","user":"Jianxiong","type":"user"},{"_id":"677272184d148b904333e874","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/5dUau7gxLk4Wm1TiiJJri.jpeg","isPro":false,"fullname":"Efstathios Karypidis","user":"Sta8is","type":"user"},{"_id":"65f7f415175d2ebe6ebaf236","avatarUrl":"/avatars/48d0fc5e06df8d1f9d7e1a40a75c0af2.svg","isPro":false,"fullname":"Jeremiah.wang","user":"JiaHWang","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2607/2607.02517.md","query":{}}">

Papers

arxiv:2607.02517

WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory

Published on Jul 2

· Submitted by

taesiri on Jul 3

Upvote

Authors:

Abstract

WorldDirector enables controllable video generation with persistent object memory by decoupling semantic motion planning from visual rendering through LLM coordination of 3D trajectories and camera movements.

Generated by Qwen/Qwen2.5-Coder-32B-Instruct

We present WorldDirector, a highly controllable video world model framework designed for persistent dynamic object memory and unrestricted viewpoint exploration. Unlike existing world models that entangle physical dynamics with pixel rendering and rely on continuous visual observation to sustain motion, our framework explicitly decouples semantic motion orchestration from visual generation. By leveraging an LLM to coordinate 3D trajectories with camera movements and subsequently employing these orchestrated trajectories as control signals for video generation, our approach ensures strict physical logic and appearance stability, successfully preserving the exact visual identities of dynamic entities even when they re-enter the scene after prolonged periods out of view. Experimental results demonstrate that our method supports the synthesis of complex and extended events with unprecedented controllability and persistent dynamic object memory. Project Page: https://worlddirector.github.io/