Fork of LongCat-Video for B200 GPU. Updates for Python, Torch, CUDA, Flash Attention 4, etc.
https://spacecruft.org/movies/LongCat-Video-B200
- Python 98.6%
- Shell 1.4%
| assets | ||
| longcat_video | ||
| .gitignore | ||
| .python-version | ||
| B200.md | ||
| download_models.sh | ||
| LICENSE | ||
| README-upstream.md | ||
| README.md | ||
| requirements-dev.txt | ||
| requirements.txt | ||
| requirements_avatar.txt | ||
| run_demo_avatar_multi_audio_to_video_25fps.py | ||
| run_demo_avatar_multi_audio_to_video_25fps.sh | ||
| run_demo_avatar_multi_audio_to_video_25fps_refine.py | ||
| run_demo_avatar_multi_audio_to_video_25fps_refine.sh | ||
| run_demo_avatar_single_audio_to_video_25fps.py | ||
| run_demo_avatar_single_audio_to_video_25fps.sh | ||
| run_demo_avatar_single_audio_to_video_25fps_refine.py | ||
| run_demo_avatar_single_audio_to_video_25fps_refine.sh | ||
| run_demo_image_to_video.py | ||
| run_demo_image_to_video.sh | ||
| run_demo_image_to_video_16fps.py | ||
| run_demo_image_to_video_16fps.sh | ||
| run_demo_image_to_video_25fps.py | ||
| run_demo_image_to_video_25fps.sh | ||
| run_demo_interactive_video.py | ||
| run_demo_interactive_video.sh | ||
| run_demo_long_video.py | ||
| run_demo_long_video.sh | ||
| run_demo_text_to_video.py | ||
| run_demo_text_to_video.sh | ||
| run_demo_video_continuation.py | ||
| run_demo_video_continuation.sh | ||
| run_streamlit.py | ||
| UPGRADE-1.5.md | ||
LongCat-Video for B200
LongCat-Video generates videos.
Fork
The original upstream code does not run on the Nvidia B200 GPU. This fork updates and optimizes the code to run on the B200. Updates include:
- Python 3.12.
- Torch 2.11.
- CUDA 13.
- Flash Attention 4.
- Add Shell script wrappers for demos.
- Use 720p, not 480p, for demo scripts.
- Add 25 FPS demo scripts.
- Add 25 FPS demo scripts for avatars with refinements.
- Model download script.
- Bug fixes.
- Cleanups.
- Formatting.
Install
Quick and dirty install, flavor to taste:
mkdir -p ~/devel
cd ~/devel/
git clone https://spacecruft.org/movies/LongCat-Video-B200
cd LongCat-Video-B200/
export CUDA_HOME=/usr/local/cuda-13.2
export PATH=$CUDA_HOME/bin:$PATH
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH
export TORCH_CUDA_ARCH_LIST="10.0"
python -m venv venv
source venv/bin/activate
pip install -U setuptools pip wheel
pip install torch==2.11.0+cu130 torchvision torchaudio torchcodec --index-url https://download.pytorch.org/whl/cu130
mkdir -p ~/devel/Dao-AILab
cd ~/devel/Dao-AILab
git clone https://github.com/Dao-AILab/flash-attention.git
cd flash-attention/flash_attn/cute/
pip install psutil
pip install -e ".[cu13]"
cd ~/devel/LongCat-Video-B200/
pip install -r requirements.txt
pip install -r requirements_avatar.txt
Upstream
Originally by the fine folks at MeiGen-AI.
See README-upstream.md for the upstream README.
Upstream license MIT.
Copyright © 2025 Meituan.
This fork is unrelated to the upstream project.
License
MIT. See LICENSE.
Copyright © 2026 Jeff Moe.
Loveland, Colorado, USA