PaintScene4D: Consistent 4D Scene Generation from Text Prompts

Vinayak Gupta¹, Yunze Man², Yu-Xiong Wang²

¹ Indian Institute of Technology Madras, ² University of Illinois Urbana-Champaign

[Project Page] [arXiv]

Overview

🎲 PaintScene4D generates photorealistic 4D scenes from text prompts which are both spatially and temporally consistent.

🚀 PaintScene4D generates 4D scenes in around 3hrs compared to SDS-based object level rendering methods that take around 12hrs to train and render.

🛠️ PaintScene4D unlocks new capabilities for camera trajectory control, an edge over traditional text-to-video model which don't have any spatial understanding.

News

• [12/06/2024] 📋 Paper is currently under review. Code will be released upon acceptance.

• [12/06/2024] 🎉 The paper appears on Arxiv.

Citation

If you find this work useful in your research, please consider citing:

@article{gupta2024paintscene4d,
      title={PaintScene4D: Consistent 4D Scene Generation from Text Prompts}, 
      author={Vinayak Gupta and Yunze Man and Yu-Xiong Wang},
      journal={https://arxiv.org/abs/2412.04471}, 
      year={2024},
}