🍓 Comfyture

Comfyture is a ready to use Docker image and compose stack designed to simplify running ComfyUI with NVIDIA GPU acceleration. Unlike traditional setups, this stack eliminates complex configuration steps while pre-installing 20+ extensions (custom nodes).

<iframe src="https://github.com/sponsors/bigbrozer/button" title="Sponsor bigbrozer" height="32" width="114" style="border: 0; border-radius: 6px;"></iframe>

Important

While functional, this project is still under active development and may have bugs during use.

Contents:

• Features
• Requirements
• Usage
  • Setting up
  • Generate your first image
• Pre-installed extensions
• Configuration
  • ComfyUI arguments
  • Envionment variables
  • Data storage
• License

Features

• Easy to run ComfyUI from Docker.
• GPU-accelerated (Nvidia CUDA only).
• Flash Attention support.
• Sage Attention support.
• Triton support.
• New Manager enabled by default.
• Do not run as root, UID/GID are configurables.
• A set of pre-installed extensions.

Warning

This project does not provide models, you must download those manually and put them in ./data/models.

Requirements

Note

This project has been tested using Archlinux BTW.

• Docker Compose

• Docker with Nvidia runtime support (run docker info):

  CDI spec directories:
  /etc/cdi
  /var/run/cdi
  Discovered Devices:
  cdi: nvidia.com/gpu=0
  cdi: nvidia.com/gpu=GPU-be3cb206-b9c5-80bf-ee15-811189491961
  cdi: nvidia.com/gpu=all
  Runtimes: nvidia runc io.containerd.runc.v2
  

Usage

Setting up

Download compose.yaml:

curl -SL -o compose.yaml https://raw.githubusercontent.com/bigbrozer/comfyture/refs/heads/main/compose.yaml

Create data directory to persist all data:

mkdir data

Adapt the compose file to your need.

Run:

docker compose up -d

Generate your first image

Open you browser on http://localhost:8188.

Follow the guide to generate your first image.

Enjoy !

Pre-installed extensions

By default, Comfyture comes with 20+ pre-installed extensions (custom nodes) to enhance your experience with ComfyUI. These extensions cover a wide range of functionalities, from advanced image processing to video generation and more.

Tip

You will see those extension as installed state from the Manager UI. There is no way to uninstall them right now but you can disable them from the Manager UI.

Here is the list of pre-installed extensions:

Extension	Description
1038lab/ComfyUI-QwenVL	Integrates the powerful Qwen-VL models.
1038lab/ComfyUI-RMBG	Advanced image background removal and precise segmentation of objects, faces, clothing, and fashion elements.
Acly/comfyui-inpaint-nodes	Inpainting nodes for ComfyUI.
Acly/comfyui-tooling-nodes	Utility and tooling nodes for ComfyUI.
city96/ComfyUI-GGUF	Nodes for working with GGUF models in ComfyUI.
cubiq/ComfyUI_essentials	Essential nodes and utilities for ComfyUI.
cubiq/ComfyUI_IPAdapter_plus	IPAdapter Plus nodes for ComfyUI.
Enemyx-net/VibeVoice-ComfyUI	Integration for using VibeVoice audio generation in ComfyUI.
Fannovel16/comfyui_controlnet_aux	Auxiliary ControlNet nodes for ComfyUI.
Fannovel16/ComfyUI-Frame-Interpolation	Video frame interpolation nodes for ComfyUI (RIFE, FILM, GMFSS Fortuna, and more).
Firetheft/ComfyUI_Local_Image_Gallery	A local image gallery node for ComfyUI.
FizzleDorf/ComfyUI_FizzNodes	A set of useful and specialized nodes for ComfyUI.
Jonseed/ComfyUI-Detail-Daemon	Adjust sigmas that generally enhance details.
kijai/ComfyUI-KJNodes	A collection of useful nodes created by kijai.
kijai/ComfyUI-segment-anything-2	Another implementation of Segment Anything nodes for ComfyUI.
kijai/ComfyUI-WanVideoWrapper	Nodes for wrapping and processing video in ComfyUI.
Kosinkadink/ComfyUI-Advanced-ControlNet	Advanced ControlNet nodes for ComfyUI.
Kosinkadink/ComfyUI-VideoHelperSuite	A suite of nodes for video processing in ComfyUI.
Lightricks/ComfyUI-LTXVideo	A collection of powerful custom nodes that extend ComfyUI's capabilities for the LTX-2 video generation model.
ltdrdata/ComfyUI-Impact-Pack	A comprehensive pack of nodes for ComfyUI.
ltdrdata/ComfyUI-Impact-Subpack	A smaller subset of nodes from the Impact pack for ComfyUI.
numz/ComfyUI-SeedVR2_VideoUpscaler	Official SeedVR2 diffusion-based video and image upscaler for ComfyUI.
PozzettiAndrea/ComfyUI-DepthAnythingV3	Nodes to support DepthAnything V3 model in ComfyUI.
pythongosssss/ComfyUI-Custom-Scripts	A collection of custom scripts for ComfyUI.
rgthree/rgthree-comfy	Nodes and tools from the rgthree ecosystem for ComfyUI.
scraed/LanPaint	Universally applicable inpainting ability for every model.
ssitu/ComfyUI_UltimateSDUpscale	Advanced upscaling nodes for Stable Diffusion in ComfyUI.
stavsap/comfyui-ollama	Ollama integration for ComfyUI.
storyicon/comfyui_segment_anything	Segment Anything model integration for ComfyUI.

Configuration

ComfyUI arguments

A default set of arguments are passed to ComfyUI by default:

--verbose
--listen=0.0.0.0
--disable-auto-launch
--multi-user
--preview-method=latent2rgb
--enable-manager

You can add extra arguments to ComfyUI. Creates extra.yaml:

# Example to enable sage attention
services:
  comfyui:
    command:
      - "--enable-sage-attention"

# Or...
# Example to enable flash attention
services:
  comfyui:
    command:
      - "--enable-flash-attention"

Then run the stack with:

docker compose -f compose.yaml -f extra.yaml up -d

If you do not like the default arguments, switch them off by setting COMFYUI_NO_DEFAULTS=true. See below.

Envionment variables

You can use these variables in compose.yaml.

Variable	Default	Description
PUID	2000	Set the User ID for files.
PGID	2000	Set the Group ID for files.
COMFYUI_NO_DEFAULTS	false	If true, disable default arguments, see how to pass your own arguments to ComfyUI.

Data storage

By default, the local ./data directory will contain everything for ComfyUI. You can customize the compose.yaml file to your needs.

Host Folder	Target Mount Point	Description
./data/python	/opt/comfyui/python	Persist Python venv & cache for fast startup
./data/custom_nodes	/opt/comfyui/app/custom_nodes	Persist custom nodes installations
./data/models	/opt/comfyui/app/models	Persist models
./data/input	/opt/comfyui/app/input	Persist input images
./data/output	/opt/comfyui/app/output	Persist output images
./data/profiles	/opt/comfyui/app/user	Persist user profiles

License

This project is licensed under the MIT License (LICENSE).

ℹ️ Note: This image uses NVIDIA's official CUDA runtime. Requires GPU hardware and NVIDIA drivers for proper operation.