🎬Show-1

David Junhao Zhang^*  Jay Zhangjie Wu^*  Jia-Wei Liu^*
Rui Zhao  Lingmin Ran  Yuchao Gu  Difei Gao  Mike Zheng Shou^✉

Show Lab, National University of Singapore
^* Equal Contribution  ^✉ Corresponding Author

---

Project Page | arXiv | PDF

News

• [10/12/2023] Code and weights released!

Setup

Requirements

pip install -r requirements.txt

Pytorch 2.0+ is highly recommended for more efficiency and speed on GPUs.

Weights

All weights are available in show lab huggingface! Please check key frames generation, interpolation, superresolution stage 1 and superresolution stage 2 modules. We also use deep-floyd-if superresolution stage 1 model for the first frame superresolution. To download deep-floyd-if models, you need follow their official instructions.

Inference

To run diffusion models for text-to-video generation, run this command:

python run_inference.py

The output videos from different modules will be stored in "outputs" folder with the gif format. The code will automatically donwload module weights from huggingface. Otherwise, you can donwload weights manually with git lfs then change the "pretrained_model_path" to your local path. Take key frames generation module for example:

git lfs install
git clone https://huggingface.co/showlab/show-1-base

Demo Video

Show-1.Demo.Video.mp4

Citation

If you make use of our work, please cite our paper.

@misc{zhang2023show1,
      title={Show-1: Marrying Pixel and Latent Diffusion Models for Text-to-Video Generation}, 
      author={David Junhao Zhang and Jay Zhangjie Wu and Jia-Wei Liu and Rui Zhao and Lingmin Ran and Yuchao Gu and Difei Gao and Mike Zheng Shou},
      year={2023},
      eprint={2309.15818},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Shoutouts

• This code heavily builds on diffusers, deep-floyd-if, modelscope, zeroscope. Thanks for open-sourcing!