comfyui — ComfyUI

Node-graph based image (and video) generation UI. More flexible than A1111 for building custom pipelines — txt2img, img2img, animation, SVD, LTX, Flux, and anything you can wire up.

Image tag

docker.io/manvarharsh/comfyui:cuda12

What’s in this image

Base: nvidia/cuda:12.4.1-cudnn-devel-ubuntu22.04
Python 3.10 (conda)
PyTorch with CUDA 12 (torch, torchvision, torchaudio)
ComfyUI (cloned at build time)
ComfyUI-Manager — install custom nodes from the UI
8 pre-installed custom nodes including:
- ComfyUI-Impact-Pack
- ComfyUI-Impact-Subpack
JupyterHub with the Podstack authenticator
OpenSSH server

Default ports

Port	Service
22	SSH
8000	JupyterHub
8188	ComfyUI WebUI

Use cases

Building reproducible image-generation pipelines via node graph
Stable Diffusion, SDXL, SD3, Flux, LTX, SVD workflows
Animation, ControlNet, IP-Adapter pipelines
Headless API rendering (--listen + REST calls into ComfyUI)

Environment variables

Variable	Description
`ENABLE_SSH`	Enable SSH server
`ENABLE_JUPYTERHUB`	Enable JupyterHub on port 8000
`ENABLE_COMFYUI`	Start ComfyUI on port 8188
`COMFYUI_PORT`	Override the listening port (default 8188)
`COMFYUI_EXTRA_ARGS`	Extra CLI args (`--highvram`, `--fp16-vae`, `--listen 0.0.0.0`)
`PODSTACK_API_URL`	Backend URL for JupyterHub token validation
`SSH_PUBLIC_KEY`	Public key for SSH

Persistence — important

Mount your storage at /data — and do not change this path.

ComfyUI’s custom nodes, models, workflows, and outputs are stored under /data. If you remap the mount point, every custom node you install will disappear on restart. Pin the mount to /data.

Typical layout:

/data/
├── models/checkpoints/    # SD/SDXL/Flux checkpoints
├── models/loras/
├── models/controlnet/
├── custom_nodes/          # nodes installed via ComfyUI-Manager
├── workflows/             # saved .json workflows
└── output/                # generated images/videos