comfyui

Why we included this project

Running ComfyUI on rented GPU instances usually means rebuilding the same environment every time you spin up a new box. This image removes that friction: it's a preconfigured ComfyUI container for NVIDIA CUDA, AMD ROCm, or CPU, so you can run the same setup on RunPod, Vast.ai, or your own hardware. The AI-Dock base layer handles authentication and a few convenience features, and environment variables give you control over the port, startup arguments, auto-update behavior, and tokens for gated model downloads from Civitai or Hugging Face. The image intentionally ships without models; you provision it with a script, and the repo includes examples for SD3 and FLUX.1, which keeps configurations reproducible across machines. That's useful for teams that want predictable, scriptable deployment of image-generation pipelines in the cloud.