↓ Skip to main content
Ellisonline.ca
Ellisonline.ca

ComfyUI - Part1

·466 words
David Ellis
Author
David Ellis
Software Industry Veteran with a clear eye on the possible.

Getting Started with ComfyUI: A Beginner’s Guide to Nodes
#

ComfyUI is a powerful, node-based graphical user interface (GUI) for Stable Diffusion that gives you total control over your AI image generation workflow. Unlike standard web interfaces, ComfyUI lets you see exactly how data flows from your text prompt to the final image. 

Here is how to set up ComfyUI and generate your very first image from text. 

Step 1: Install ComfyUI
#

Getting ComfyUI up and running is straightforward, especially if you are on Windows. 

  1. Download the Architecture: Go to the official ComfyUI GitHub repository and download the Windows portable standalone build.
  2. Extract the Files: Extract the downloaded .7z file using a tool like 7-Zip to a folder on your drive (preferably an SSD).
  3. Download a Model: You need a checkpoint model to generate images. Download a popular model like SDXL 1.0 or Stable Diffusion 1.5 from Civitai or Hugging Face.
  4. Place the Model: Move your downloaded model file (ending in .safetensors) into the directory: ComfyUI_windows_portable/ComfyUI/models/checkpoints/
  5. Launch: Double-click run_nvidia_gpu.bat (or run_cpu.bat if you do not have a dedicated Nvidia graphics card).

Step 2: Understand the Default Workflow
#

When ComfyUI opens in your browser, you will see a default layout of interconnected blocks called Nodes. 

[Load Checkpoint] —> [CLIP Text Encode (Prompt)] —> [Ksampler] —> [VAE Decode] —> [Save Image]

  • Load Checkpoint: Chooses your AI model.
  • CLIP Text Encode: Where you type your positive and negative prompts.
  • KSampler: The engine that turns random noise into an image based on your prompt.
  • VAE Decode: Translates the hidden mathematical image data into actual pixels you can see.
  • Save Image: Displays and saves your final file.

Step 3: Generate Your First Image
#

  1. Select Your Model: In the Load Checkpoint node, click the dropdown menu and select the model you placed in your checkpoints folder.
  2. Enter Your Prompt: 
  • Locate the top CLIP Text Encode node (connected to the positive slot on the KSampler) and type: A futuristic cyberpunk city at night, neon lights, highly detailed, 8k resolution.
  • Locate the bottom CLIP Text Encode node (connected to the negative slot) and type: blurry, low quality, distorted.
  1. Queue the Prompt: Look at the control panel on the right side of the screen and click Queue Prompt (or press Ctrl + Enter).

You will see a green border highlight each node as ComfyUI processes the data. Within a few seconds, your generated masterpiece will appear inside the Save Image node! 

Why Use ComfyUI?
#

While the node interface looks intimidating at first, it is highly optimized. ComfyUI uses significantly less VRAM than other interfaces and allows you to save your entire workflow layout directly inside the images you generate. To recreate any image in the future, simply drag and drop the saved PNG file right back into the ComfyUI browser tab!