Published on

AMD GPU run Fooocus on Windows - step by step guide


"Fooocus is a rethinking of Stable Diffusion and Midjourney designs"

What is Fooocus?

Fooocus is an innovative software built on the foundation of Gradio. It's a groundbreaking approach to image generation, combining the offline, open-source, and free nature of Stable Diffusion with the simplified, intuitive interface of Midjourney. This unique blend offers a straightforward, hassle-free experience for users, focusing on creativity rather than technical complexity.

Key Features of Fooocus:

  • Offline and Open Source: Inspired by Stable Diffusion, Fooocus operates offline and is completely open-source, ensuring accessibility and transparency.
  • Ease of Use: Learning from Midjourney, Fooocus eliminates the need for manual tweaking. Users can concentrate on crafting prompts and visualizing outcomes without getting bogged down in technical details.
  • Automated Optimizations: The software includes numerous internal optimizations and quality enhancements, all automated to save you time and effort.
  • Simple Installation: Fooocus boasts an incredibly user-friendly installation process. From clicking "download" to generating your first image, you'll need less than three mouse clicks.

A Word of Caution

In the age of information overload, it's crucial to source your software from reliable channels. Recently, numerous fake websites have surfaced claiming to offer Fooocus. Be wary of these imitations. For a genuine Fooocus experience, ensure you visit the official source.

Windows Installation Guide for AMD GPU users

Video tutorial

The following has been tested on an AMD 6700XT GPU on Windows 10. It should be working on all recent generation AMD GPUs, even on APU (as in my previous article, APU can be turned into 16GB VRAM GPU), on Windows 10 and 11.

I recommend a system RAM of at least 32 GB.

The Git and Ananconda/Miniconda (for python virtual environment) are needed. Please make sure to install them first (for video tutorial, see this video).

On Windows search box, search and open 'Anaconda prompt', a terminal will be opened.

Step 1: create a Python virtual environment. I used python 3.10.12.

conda create -n FooocusEnv python=3.10.12 -y

# activate it

conda activate FooocusEnv

The following steps are all within the python virtual environment.

Step 2: clone the Fooocus code to local PC and go into the directory.

git clone https://github.com/lllyasviel/Fooocus.git
cd Fooocus

Step 3: install dependencies

pip install torch-directml
pip install -r requirements_versions.txt

Step 4: Start the Fooocus program and open in browser. It will download some stable diffusion model files (about 6.6 GB). You can also download your custom models manually and put them in Fooocusmodelscheckpoints.

python entry_with_update.py --directml

The web page UI should be opened automatically in your default browser. If not, please type in your favorite browser. you should see the UI similar to this:


I was using 6700XT GPU. Let's give it a try!

To get better speed, choose the 'Extreme speed' mode in advanced->Setting (see below):

This is the information shown in terminal. It took about 100 seconds for one image.

If you want better quality images, choose the 'Speed' mode in advanced->Setting:

The logs are in terminal. It needs more steps (30) and significant more time (500 seconds).

If you want to get the most out of your AMD GPU, I recommend to use Linux OS. Please check out some of my videos.

Stay Connected for More Adventures!

If you've enjoyed this article, don't let it end here! I'm constantly exploring new horizons, uncovering hidden gems, and sharing insightful tips and tricks. By following my blog, you'll be the first to dive into fresh, engaging content that sparks your curiosity and fuels your passion. So, make sure to hit that follow button, and let's continue this adventure together. Your support means the world to me, and it's what keeps this blog thriving. Here's to many more exciting discoveries ahead!

Please also support and subscribe to my youtube channel and follow me on X(Twitter) https://twitter.com/TechPractice1for latest updates!