LAMP：语言辅助的运动规划用于可控视频生成
LAMP: Language-Assisted Motion Planning for Controllable Video Generation
用 LLM 做程序合成，把自然语言解析成显式 3D 轨迹（物体运动 + 相机轨迹），再作为视频生成的运动条件；这种「语言→结构化运动程序→确定性轨迹」的中间表示，可复用于任何需要精确运动控制的可控视频生成或相机运镜合成。
原文依据 leverages large language models (LLMs) as motion planners to translate natural language descriptions into explicit 3D trajectories
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论视频生成在视觉保真度和可控性方面已取得显著进展，能够以文本、布局或运动作为条件。其中，运动控制——即指定物体动态与相机轨迹——对于构建复杂的电影化场景至关重要，但现有接口仍然有限。我们提出 LAMP，利用大语言模型（LLM）作为运动规划器，将自然语言描述转换为动态物体与（相对定义的）相机的显式 3D 轨迹。LAMP 定义了一种受电影摄影惯例启发的运动领域特定语言（DSL）。借助 LLM 的程序合成能力，LAMP 从自然语言生成结构化运动程序，并将其确定性地映射为 3D 轨迹。我们构建了一个大规模程序化数据集，将自然文本描述与相应的运动程序和 3D 轨迹配对。实验表明，与最优的替代方案相比，LAMP 在运动可控性以及与用户意图的对齐方面表现更优，建立了首个直接从自然语言规范生成物体运动与相机运动的框架。视频与代码见项目主页 https://cyberiada.github.io/LAMP/ 。
英文原文摘要 1297 字符
Video generation has achieved remarkable progress in visual fidelity and controllability, enabling conditioning on text, layout, or motion. Among these, motion control -- specifying object dynamics and camera trajectories -- is essential for composing complex, cinematic scenes, yet existing interfaces remain limited. We introduce LAMP, that leverages large language models (LLMs) as motion planners to translate natural language descriptions into explicit 3D trajectories for dynamic objects and (relatively defined) cameras. LAMP defines a motion domain-specific language (DSL), inspired by cinematography conventions. By harnessing program synthesis capabilities of LLMs, LAMP generates structured motion programs from natural language, which are deterministically mapped to 3D trajectories. We construct a large-scale procedural dataset pairing natural text descriptions with corresponding motion programs and 3D trajectories. Experiments demonstrate LAMP's improved performance in motion controllability and alignment with user intent compared to state-of-the-art alternatives, establishing the first framework for generating both object and camera motions directly from natural language specifications. Video and code are available at the project page at https://cyberiada.github.io/LAMP/ .