Skip to main content

ComfyUI manual

RyanAbout 2 min

ComfyUI: The Ultimate Guide to Stable Diffusion's Powerful and Modular GUI

Welcome to the comprehensive, community-maintained documentation for ComfyUIopen in new window, the cutting-edge, modular Stable Diffusion GUI and backend. This guide is designed to help you quickly get started with ComfyUI, run your first image generation, and explore advanced features.

Quick Start: Installing ComfyUI

For the most up-to-date installation instructions, please refer to the official ComfyUI GitHub READMEopen in new window. Choose the section relevant to your operating system (Linux, macOS, or Windows) for tailored instructions.

Essential First Step: Downloading a Stable Diffusion Model

New to Stable Diffusion? Your first task is to acquire a model checkpoint for image generation.

"Pro Tip for Experienced Users"

Already have model files? Skip the download! Rename extra_model_paths.yaml.example to extra_model_paths.yaml, edit the file to point to your existing models, and restart ComfyUI.

For beginners, we recommend exploring popular model repositories:

  1. CivitAIopen in new window - A vast collection of community-created models
  2. HuggingFaceopen in new window - Home to numerous official and fine-tuned models

Download your chosen model checkpoint and place it in the models/checkpoints directory (create it if needed). Restart ComfyUI to load your new model.

Getting Started: Your First ComfyUI Workflow

With ComfyUI running in your browser, you're ready to begin. Here's how to navigate and use the interface:

  • Canvas Navigation: Drag the canvas or hold ++space++ and move your mouse
  • Zoom: Use your mouse scroll wheel
  • Reset Workflow: Click Load Default in the menu if you need a fresh start
ComfyUI Default Workflow
Explore ComfyUI's default startup workflow (click for full-size view)

Optimizing Your Workflow: Quick Preview Setup

Before running your first generation, let's modify the workflow for easier image previewing:

  1. Remove the Save Image node (right-click and select Remove)
  2. Add a PreviewImage node (double-click canvas, type "preview", select PreviewImage)
  3. Connect the IMAGE output from VAE Decode to the images input of PreviewImage

This setup allows you to preview images without automatic saving, streamlining your workflow.

Generating Your First Image

Ready to create? Simply click Queue Prompt in the menu or use the keyboard shortcut Cmd+Enter (Mac) / Ctrl+Enter (Windows/Linux).

Advanced Feature: Loading External Workflows

ComfyUI, like many Stable Diffusion interfaces, embeds workflow metadata in generated PNGs. This feature enables easy sharing and reproduction of complex setups.

To load a workflow from an image:

  1. Click the Load button in the menu
  2. Or drag and drop the image into the ComfyUI window

The associated workflow will automatically load, complete with all nodes and settings.

"Troubleshooting: No Workflow Displayed"

If loading an image doesn't reveal a workflow, the metadata may have been stripped. Contact the image creator for an unmodified version.

Expanding Your ComfyUI Journey

This guide has equipped you with the essentials to start your ComfyUI adventure. The node-based interface allows for intricate, customizable image generation pipelines limited only by your creativity.

As you progress, you'll likely have questions about specific nodes, advanced techniques, and how to achieve particular effects. The following sections of our documentation are designed to answer these questions and deepen your understanding of ComfyUI's capabilities.

Explore, experiment, and enjoy the vast possibilities of AI-powered image generation with ComfyUI!