Skip to main content
←EXPLORE MORE
AI
BEGINNER
8 MIN

Flux 1 Dev: Installation Tutorial for ComfyUI

A structured step-by-step guide for Flux 1 Dev with direct download links and a desktop workflow.

Flux 1 Dev: Installation Tutorial for ComfyUI
Flux 1 Dev: Installation Tutorial for ComfyUI

In 2026, ComfyUI can be installed without terminals or manual Python setup. The official Desktop App handles dependencies, updates, and includes ComfyUI-Manager by default.

This guide follows a clear structure and stays fully graphical from download to first image.

Quickstart

  • Install the ComfyUI Desktop App
  • Download the Flux 1 Dev model files and place them
  • Load a workflow and run your first prompt

Why the Desktop App?

  • No command line, Git, or pip
  • Automatic Python and PyTorch setup
  • ComfyUI-Manager included
  • Clean, isolated installation

Requirements

  • Windows: 10 or 11 (macOS works similarly)
  • GPU: NVIDIA recommended
  • VRAM: at least 8 GB, ideally 12-16 GB

Step 1: Install ComfyUI Desktop

Official download page: https://www.comfy.org/download

  1. Select Windows (or macOS)
  2. Download the installer
  3. Run it and keep the default install path
  4. Let the installer fetch dependencies (5-15 minutes)

After install, ComfyUI appears in the Start Menu or on the desktop.

Step 2: First Launch

On first start, the app:

  • Launches the ComfyUI server
  • Opens your browser at http://127.0.0.1:8188

If no tab opens, paste the URL into your browser. The Manager button should be visible on the right.

Step 3: Download Flux 1 Dev Models

Main model (models/diffusion_models/):

Text encoders (models/clip/):

VAE (models/vae/):

Download the files from Hugging Face.

Step 4: Place Models in the Right Folders

Default models folder:

C:\Users\YourName\AppData\Local\Programs\ComfyUI\models

Create these subfolders if needed:

  • clip
  • vae
  • diffusion_models (or unet)

Place the files:

Text encoders (models/clip/):

  • clip_l.safetensors
  • t5xxl_fp8_e4m3fn.safetensors

VAE (models/vae/):

  • ae.safetensors

Main model (models/diffusion_models/):

  • flux1-dev-fp8.safetensors

Drag the files into the matching folders. Restart ComfyUI afterward so the model lists refresh.

Step 5: Load a Flux Workflow

Examples: https://comfyanonymous.github.io/ComfyUI_examples/flux/

  1. Pick a Flux Dev text-to-image workflow
  2. Drag the example image or JSON file into the ComfyUI canvas

Step 6: Configure the Workflow

Diffusion model:

  • Select flux1-dev-fp8.safetensors
  • Set weight_dtype to fp8 if available

CLIP / DualCLIP Loader:

  • clip_l.safetensors
  • t5xxl_fp8_e4m3fn.safetensors

VAE Loader:

  • ae.safetensors

Recommended starting values:

  • Steps: 20-50
  • Guidance: about 3.5
  • CFG: 1.0

Step 7: Run a Prompt

Enter a clear, specific prompt and click Queue Prompt. The output appears in the app's default output folder.

Notes

  • The Desktop App keeps ComfyUI updated.
  • Missing nodes can be installed via the Manager panel.
  • FP8 models work well with 12-16 GB VRAM.
  • If models do not show up, restart the app.

Troubleshooting

—Models not listed

  • Verify folder paths
  • Restart the app

—Out of memory

  • Lower resolution (e.g., 768 x 768)
  • Reduce Steps
  • Run again

Conclusion

Flux 1 Dev runs locally without a terminal or manual Python setup. Once the models are in the right folders, the workflow is stable and repeatable.

FAQ

—Do I need Git or pip?

No. The Desktop App installs everything automatically.

—Which files are required?

One main model, two text encoders, and one VAE. The four files above are the minimal set.