WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory - 深度解析 论文来源:ArXiv (2607.02517) 作者:Hanlin Wang, Hao Ouyang, Qiuyu Wang, Wen Wang, Qingyan Bai 等 分类:cs.CV 发布时间:2026-07-02T17:59:59Z 解读时间:2026年07月06日 09:04:46 📋 论文基本信息 标题:WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory 作者:Hanlin
论文来源:ArXiv (2607.02517)
作者:Hanlin Wang, Hao Ouyang, Qiuyu Wang, Wen Wang, Qingyan Bai
等
分类:cs.CV
发布时间:2026-07-02T17:59:59Z
解读时间:2026年07月06日 09:04:46
标题:WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory
作者:Hanlin Wang, Hao Ouyang, Qiuyu Wang, Wen Wang, Qingyan Bai
ArXiv ID:2607.02517
链接:https://arxiv.org/abs/2607.02517v1
分类:cs.CV
研究领域:3D视觉
本论文研究了 3D视觉 领域的重要问题。
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
该研究对于解决当前领域面临的挑战具有重要意义。
论文提出了一种新颖的方法来解决相关问题。
论文通过大量实验验证了所提方法的有效性。
本论文的主要创新点包括:
该方法在 3D视觉 领域具有广阔的应用前景。
建议读者根据自身需求深入阅读相关文献。
本论文为相关研究做出了重要贡献。
本文由 AI 自动生成。要启用 Qwen 深度分析,请配置 DASHSCOPE_API_KEY。