license: other
license_name: license
license_link: https://github.com/TQTQliu/Light-X/blob/master/LICENSE
tags:
- computer-vision
- video-generation
- video-editing
- 4d
- relighting
- camera-control
Light-Syn Dataset
📄 Paper | 🚀 Project Page | 💻 GitHub
Introduction
Light-Syn is a synthetic dataset designed for learning controllable video generation under varying camera viewpoints and illumination conditions.
It is constructed via a synthetic data generation pipeline that produces paired video samples with controlled viewpoint and lighting variations from monocular visual inputs, and is introduced as training data for the Light-X framework, which enables joint control of camera motion and illumination in generative video models.
Release Note
The current public release of Light-Syn includes only the static scene subset, due to copyright and data usage constraints.
For access to the complete dataset (e.g., dynamic scenes), please contact Tianqi Liu (tq_liu at hust.edu.cn).
Dataset Structure
The dataset is organized as a collection of per-sample folders, where each folder corresponds to one scene and contains paired videos, text prompts, and lighting conditions.
A high-level directory structure is as follows:
Light-Syn/
├── scene0/
│ ├── input.mp4 # Input video
│ ├── target.mp4 # Target video under new view / lighting
│ ├── render.mp4 # Rendered result
│ ├── mask.mp4 # Rendered valid-region mask
│ ├── relit_render.mp4 # Relit rendering result
│ ├── prompt.txt # Text prompt describing the scene
│ ├── id.txt # Reference frame index (integer)
│ ├── ref_light.png # Reference lighting image (required for reference-based relighting)
│ └── scene_000.exr # HDR environment map (required for HDR-based relighting)
├── scene1/
│ └── ...
Citation
If you find our work useful for your research, please consider citing our paper:
@article{liu2025light,
title={Light-X: Generative 4D Video Rendering with Camera and Illumination Control},
author={Liu, Tianqi and Chen, Zhaoxi and Huang, Zihao and Xu, Shaocong and Zhang, Saining and Ye, Chongjie and Li, Bohan and Cao, Zhiguo and Li, Wei and Zhao, Hao and others},
journal={arXiv preprint arXiv:2512.05115},
year={2025}
}