README.md

---
license: apache-2.0
language:
- en
---
arxiv.org/abs/2503.15667



<p align="center">

  <h2 align="center">[CVPR'25]DiffPortrait360: Consistent Portrait Diffusion for 360 View Synthesis</h2>
  <p align="center">
      <a href="https://freedomgu.github.io/">Yuming Gu</a><sup>1,2</sup>
      ·
      <a href="https://p0lyfish.github.io/portfolio/">Phong Tran</a><sup>2</sup>
    ·  
      <a href="https://paulyzheng.github.io/about/">Yujian Zheng</a><sup>2</sup>
    ·  
      <a href="https://hongyixu37.github.io/homepage/">Hongyi Xu</a><sup>3</sup>
    ·  
      <a href="https://lhyfst.github.io/">Heyuan Li</a><sup>4</sup>
    ·  
      <a href="https://www.linkedin.com/in/adilbek-karmanov?originalSubdomain=ae">Adilbek Karmanov</a><sup>2</sup>
    ·  
      <a href="https://hao-li.com">Hao Li</a><sup>2,5</sup>
    <br>
    <sup>1</sup>Unviersity of Southern California &nbsp;<sup>2</sup>MBZUAI &nbsp; <sup>3</sup>ByteDance Inc. &nbsp; 
    <br>
    <sup>4</sup>The Chinese University of Hong Kong, Shenzhen&nbsp; <sup>5</sup>Pinscreen Inc.
    <br>
    </br>
        <a href="https://arxiv.org/abs/2503.15667">
        <img src='https://img.shields.io/badge/arXiv-diffportrait360-green' alt='Paper PDF'>
        </a>
        <a href='https://freedomgu.github.io/DiffPortrait360/'>
        <img src='https://img.shields.io/badge/Project_Page-diffportrait360-blue' alt='Project Page'></a>
        <a href='https://huggingface.co/gym890/diffportrait360'>
        <img src='https://img.shields.io/badge/%F0%9F%A4%97%20HuggingFace-Model-yellow'></a>
     </br>
</p>




## 📜 Requirements
* An NVIDIA GPU with CUDA support is required. 
  * We have tested on a single A6000 GPU.
  * **Minimum**: The minimum GPU memory required is 30GB for generating a single NVS video (batch_size=1) of 32 frames each time.
  * **Recommended**: We recommend using a GPU with 40GB of memory.
* Operating system: Linux


## 🧱 Download Pretrained Models

```bash
Diffportrait360
|----...
|----pretrained_weights
  |----back_head-230000.th # back head generator
  |----model_state-3400000.th # diffportrait360 main module
  |----easy-khair-180-gpc0.8-trans10-025000.th
|----...
``` 

## 🔗 BibTeX
If you find [Diffportrait360](https://arxiv.org/abs/2503.15667) is useful for your research and applications, please cite Diffportrait360 using this BibTeX:

```BibTeX
@article{gu2025diffportrait360,
  title={DiffPortrait360: Consistent Portrait Diffusion for 360 View Synthesis},
  author={Gu, Yuming and Tran, Phong and Zheng, Yujian and Xu, Hongyi and Li, Heyuan and Karmanov, Adilbek and Li, Hao},
  journal={arXiv preprint arXiv:2503.15667},
  year={2025}
}
```


## License

Our code is distributed under the Apache-2.0 license.


## Acknowledgements

This work is supported by the Metaverse Center Grant from the MBZUAI Research Office. We appreciate the contributions from [Diffportrait3D](https://github.com/FreedomGu/DiffPortrait3D), [PanoHead](https://github.com/SizheAn/PanoHead), [SphereHead](https://lhyfst.github.io/spherehead/), [ControlNet](https://github.com/lllyasviel/ControlNet) for their open-sourced research. We thank [Egor Zakharov](https://egorzakharov.github.io/), [Zhenhui Lin](https://www.linkedin.com/in/zhenhui-lin-5b6510226/?originalSubdomain=ae), [Maksat Kengeskanov](https://www.linkedin.com/in/maksat-kengeskanov/%C2%A0/), and Yiming Chen for the early discussions, helpful suggestions, and feedback.