This repository provides Docker images and startup scripts for ComfyUI, a node-based GUI for Stable Diffusion. It simplifies the setup and deployment of ComfyUI, making it accessible for beginners and users who prefer containerized environments, particularly for NVIDIA GPU users.

How It Works

The project offers pre-built Docker images for various CUDA versions (12.1, 12.4) and configurations, including slim versions with minimal dependencies and "megapak" versions with development kits. It also provides ROCm support for AMD GPUs and a nightly build for preview versions. The images are designed for ease of use, with options for low-privilege or root users within the container, facilitating deployment on systems like WSL2.

Quick Start & Requirements

• NVIDIA GPU:

  mkdir -p storage
  docker run -it --rm \
    --name comfyui-cu124 \
    --gpus all \
    -p 8188:8188 \
    -v "$(pwd)"/storage:/root \
    -e CLI_ARGS="" \
    yanwk/comfyui-boot:cu124-slim
  

• Prerequisites: Docker, NVIDIA Container Toolkit (for NVIDIA GPUs), ROCm (for AMD GPUs).
• Resources: Requires sufficient disk space for the image and models.
• Documentation: Docker Hub, Chinese README

Highlighted Details

• Offers multiple image variants (e.g., cu121, cu124-slim, cu121-megapak, rocm, nightly).
• Includes ComfyUI-Manager in some variants for easier extension management.
• Provides a cu124-cn image with mirror sites for users in mainland China.
• Dedicated image (comfy3d-pt25) for ComfyUI-3D-Pack.

Maintenance & Community

• The project is actively maintained, with various image tags indicating different CUDA versions and feature sets.
• Links to Chinese documentation are provided.

Licensing & Compatibility

• License: Mulan Public License, Version 2.
• Compatibility: Suitable for commercial use and linking with closed-source applications, as the license is permissive.

Limitations & Caveats

The cu121 and cu124-slim images are not recommended for Podman or rootless deployments, suggesting alternative images for those use cases.