Our pipeline does not rely on any category-specific prior or dynamic structural annotations, and therefore can be applied to generate 4D scenes of a wide range of interactions of dynamic objects.
Our pipeline does not rely on any category-specific prior or dynamic structural annotations, and therefore can be applied to generate 4D scenes of a wide range of interactions of dynamic objects.
Text prompt: "A man closing the lid of a laptop."
Select Scene:
Select Scene:
Select Scene:
@misc{lyu2026chord,
title={Choreographing a World of Dynamic Objects},
author={Yanzhe Lyu and Chen Geng and Karthik Dharmarajan and Yunzhi Zhang and Hadi AlZayer and Shangzhe Wu and Jiajun Wu},
year={2026},
eprint={2601.04194},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2601.04194},
}