Local AI Local AI Foundations

First Image with ComfyUI

Install ComfyUI and generate your first image

62 of 66

Install ComfyUI

ComfyUI is a node-based interface for Stable Diffusion and Flux. Official site: comfy.org

Option 1: Portable (Windows)

Download the portable release from comfyanonymous.github.io. Extract and run 

run_nvidia_gpu.bat

Option 2: Git clone

git clone https://github.com/comfyanonymous/ComfyUI.git
cd ComfyUI
pip install -r requirements.txt
python main.py

Download a base model

Place a checkpoint in:

ComfyUI/models/checkpoints/

Good starters:

  • SDXL Base for general use
  • Flux.1 [dev] for high quality

Default workflow

When ComfyUI opens, you see a node graph:

Load Checkpoint → CLIP Text Encode → KSampler → VAE Decode → Save Image
  1. Select your checkpoint in Load Checkpoint
  2. Enter a positive prompt in the top CLIP Text Encode
  3. Enter a negative prompt in the bottom one
  4. Click Queue Prompt

First prompt

a small robot reading a book in a sunny library, digital art

Common first-run errors

  • CUDA out of memory: reduce image size or use a smaller model
  • Model not found: check the checkpoint path
  • Black output: ensure the VAE is loaded correctly

Working out which model to run this on? See The Codex. Packaging it as a reusable skill? See The Armory.