Overview
Research area: cs.CV
Authors: Hanlin Wang, Hao Ouyang, Qiuyu Wang, Wen Wang, Qingyan Bai, Ka Leong Cheng, Yue Yu, Yixuan Li, Yihao Meng, Zichen Liu, Yanhong Zeng, Yujun Shen, Qifeng Chen
Published: 2026-07-02
arXiv: 2607.02517
Abstract (translation)
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.
Key ideas
- Decoupled architecture: semantic motion orchestration is separated from visual generation, rather than entangling physical dynamics with pixel rendering.
- LLM-based orchestration: an LLM coordinates 3D object trajectories alongside camera movements.
- Trajectory-conditioned generation: the orchestrated trajectories act as control signals for the video generation model.
- Persistent dynamic memory: objects retain their exact visual identities even after long absences from view.
- Free viewpoint exploration: the framework supports unrestricted camera control while maintaining physical logic and appearance stability.
*Auto-collected on 2026-08-28.*