Avatarstudio

<p align="center"> <h2 align="center">AvatarStudio: High-fidelity and Animatable 3D Avatar Creation from Text</h2> <p align="center"> <a href="http://jeff95.me/"><strong>Jianfeng Zhang*</strong></a> · <a href="https://scholar.google.com.sg/citations?user=8gm-CYYAAAAJ&hl=en"><strong>Xuanmeng Zhang*</strong></a> · <a href="https://scholar.google.com/citations?user=ZfUJGskAAAAJ&hl=zh-CN/"><strong>Huichao Zhang</strong></a> · <a href="https://scholar.google.com.sg/citations?user=8gm-CYYAAAAJ&hl=en"><strong>Jun Hao Liew</strong></a> · <a href="https://zhangchenxu528.github.io/"><strong>Chenxu Zhang</strong></a> · <a href="https://scholar.google.com.sg/citations?user=RMSuNFwAAAAJ&hl=en"><strong>Yi Yang</strong></a> · <a href="https://sites.google.com/site/jshfeng/home"><strong>Jiashi Feng</strong></a> <br> <br> <a href="assets/preprint/AvatarStudio.pdf"><img src='https://img.shields.io/badge/Paper-AvatarStudio-red' alt='Paper PDF'></a> <a href="https://arxiv.org/abs/2311.17917"><img src='https://img.shields.io/badge/arXiv-AvatarStudio-blue' alt='Paper PDF'></a> <a href='http://jeff95.me/projects/avatarstudio.html'><img src='https://img.shields.io/badge/Project_Page-AvatarStudio-green' alt='Project Page'></a> </p> <table align="center"> <tr> <td align="center" colspan="2"><b>3D Avatar Creation</b></td> </tr> <tr> <td> <img src="https://github.com/magic-research/avatarstudio/assets/25397555/a3cd9b3d-55d0-438b-ae03-6f850f3d7eea"> </td> <td> <img src="https://github.com/magic-research/avatarstudio/assets/25397555/4f1f4780-ff2c-4440-82a9-bb0aafb0056b"> </td> </tr> <tr> <td> <img src="https://github.com/magic-research/avatarstudio/assets/25397555/b6192eec-d41a-45c9-ba71-6b496a26ce62"> </td> <td> <img src="https://github.com/magic-research/avatarstudio/assets/25397555/4aee3b7b-0ba2-4787-91b8-911ddd9696f8"> </td> </tr> <tr> <td align="center" colspan="2"><b>Stylized Avatar Generation</b></td> </tr> <tr> <td> <img src="https://github.com/magic-research/avatarstudio/assets/25397555/9ebc8aca-69a1-4a73-8fd3-168523e5147a"> </td> <td> <img src="https://github.com/magic-research/avatarstudio/assets/25397555/48223a1f-15fe-4d87-8a15-7b42c1efe3e7"> </td> </tr> <tr> <td> <img src="https://github.com/magic-research/avatarstudio/assets/25397555/a0c4a4b9-65e8-4023-aaf5-d7486d48693a"> </td> <td> <img src="https://github.com/magic-research/avatarstudio/assets/25397555/7602a373-9f87-43fc-a771-eac465a16649"> </td> </tr> <td align="center" colspan="2"><b>Multimodal Avatar Animation</b></td> <tr> <td> <img src="https://github.com/magic-research/avatarstudio/assets/25397555/63957228-971f-4c6b-91fc-a49cf83c5861"> </td> <td> <img src="https://github.com/magic-research/avatarstudio/assets/25397555/4d2bdfa7-e6ef-4c50-9a60-64097eab3eec"> </td> </tr> </table>

Introducing AvatarStudio, a generative model for high-fidelity and animatable human avatar creation from text. It supports various applications, including stylized avatar generation and multimodal animation.

📢 News

• [2023-11-30]: Repo created. Code will come soon. Stay tuned! Check our project page here.

<!-- ## ⚒️ Requirements * We recommend Linux for performance and compatibility reasons. * 1 high-end NVIDIA GPU. We have done all testing and development using V100 GPUs. * Install `Python >= 3.8` and `PyTorch >= 1.12`. We have tested on `torch1.12.1+cu113, but other versions should also work fine. * (Optional, Recommended) Install ninja to speed up the compilation of CUDA extensions:```pip install ninja``` * Install dependencies: ```pip install -r requirements.txt ``` ## 🏃‍♂️ Getting Started #### Clone the gitlab code and necessary files: ```bash git clone https://github.com/magic-research/avatarstudio.git ``` --> <!-- ## 🙀 Train the model You can train new models using `launch.py`. For example: ```bash python3 launch.py --config configs/avatarstudio.yaml --train --gpu 0 system.prompt_processor.prompt="Captain America, Marvel Character" ``` ## 🙉 Inference You can generate the multi-view visualization with `launch.py`. For example: ```bash python3 launch.py --config /path/to/trial/dir/configs/parsed.yaml --test --gpu 0 resume=path/to/trial/dir/ckpts/last.ckpt ``` You can specify `--img_res` to be the image resolution and `--resume_pretrained` to be the path of checkpoints. -->

🙏 Credit

AvatarStudio builds upon several previous works:

• threestudio
• GET3D
• GETAvatar

We would like to thank the authors for their contribution to the community!

🎓 Citation

If you find our work useful, please consider citing:

@inproceedings{zhang2023avatarstudio,
    title={AvatarStudio: High-fidelity and Animatable 3D Avatar Creation from Text},
    author={Zhang, Jianfeng and Zhang, Xuanmeng and Zhang, Huichao and Liew, Jun Hao and Zhang, Chenxu and Yang, Yi and Feng, Jiashi},
    booktitle={arXiv},
    year={2023}
}