Spatia

<div align="center"> <h1 align="center"> <span style="color: #4F46E5; font-size: 1.2em;">Spatia</span>: Video Generation with Updatable Spatial Memory </h1> <p align="center" style="font-size: 1.1em; color: #555;"> <strong>Long-horizon, spatially consistent video generation enabled by persistent 3D scene point clouds and dynamic-static disentanglement.</strong> </p> <div align="center"> <a href="https://github.com/ZhaoJingjing713">Jinjing Zhao</a>^*1&nbsp;&nbsp; <a href="https://scholar.google.com/citations?user=-ncz2s8AAAAJ">Fangyun Wei</a>^*2&nbsp;&nbsp; <a href="https://www.liuzhening.top">Zhening Liu</a>³&nbsp;&nbsp; <a href="https://hongyanz.github.io/">Hongyang Zhang</a>⁴&nbsp;&nbsp; <a href="http://changxu.xyz/">Chang Xu</a>¹&nbsp;&nbsp; <a href="https://scholar.google.com/citations?user=djk5l-4AAAAJ">Yan Lu</a>² </div> <p align="center" style="font-size: 0.9em; color: #666;"> ¹The University of Sydney&nbsp;&nbsp; ²Microsoft Research&nbsp;&nbsp; ³HKUST&nbsp;&nbsp; ⁴University of Waterloo <br> <small>^*Equal Contribution</small> </p> <p align="center"> <a href="https://arxiv.org/pdf/2512.15716"> <img src="https://img.shields.io/badge/arXiv-Paper-B31B1B?style=flat&labelColor=555555&logo=arxiv&logoColor=white" alt="arXiv"> </a> &nbsp; <a href="https://zhaojingjing713.github.io/Spatia/"> <img src="https://img.shields.io/badge/Project-Page-4F46E5?style=flat&labelColor=555555&logo=googlechrome&logoColor=white" alt="Project Page"> </a> &nbsp; <a href="https://github.com/ZhaoJingjing713/Spatia"> <img src="https://img.shields.io/badge/Codes-FF5722?style=flat&labelColor=555555&logo=github&logoColor=white" alt="Codes"> </a> </p> </div>

---

📖 Abstract

Existing video generation models struggle to maintain long-term spatial and temporal consistency due to the dense, high-dimensional nature of video signals. To overcome this limitation, we propose Spatia, a spatial memory-aware video generation framework that explicitly preserves a 3D scene point cloud as persistent spatial memory.

Spatia iteratively generates video clips conditioned on this spatial memory and continuously updates it through visual SLAM. This dynamic-static disentanglement design enhances spatial consistency throughout the generation process while preserving the model's ability to produce realistic dynamic entities.

Furthermore, Spatia enables applications such as:

• Explicit Camera Control
• 3D-Aware Interactive Editing
• Long-horizon Scene Exploration

<br> <div align="center"> <img src="./teaser.png" width="100%" alt="Spatia Teaser"/> </div>

---

Citation

If you find this project useful, please cite the paper.

@inproceedings{zhao2026spatia,
  title={Spatia: Video Generation with Updatable Spatial Memory},
  author={Zhao, Jinjing and Wei, Fangyun and Liu, Zhening and Zhang, Hongyang and Xu, Chang and Lu, Yan},
  booktitle={Proceedings of the IEEE/cvf conference on computer vision and pattern recognition},
  year={2026}
}

---

<p align="center"> <small>© 2025 Spatia Project. Licensed under <a href="http://creativecommons.org/licenses/by-sa/4.0/">CC BY-SA 4.0</a>.</small> </p>