6 mins read

Fooocus: The Easiest Way To Run Stable Diffusion With Impressive Results – Guide

Are you interested in running Stable Diffusion XL on your PC but want to avoid complex procedures? Look no further than Fooocus, the perfect blend of simplicity and power for generating high-quality images. Whether you’re aiming to create vibrant stock photos or exquisite paintings, Fooocus streamlines the process, making it accessible and straightforward for users of all levels. Let’s have a look in this simple guide.

What is Fooocus?

Fooocus is one of the easiest way to use Stable Diffusion. It’s really simple. Unlike other tools like ComfyUI or stable-diffusion-webui, it lets you make great images even with just a single prompt. It’s designed to work on PCs with 8GB VRAM GPUs, but if you’ve got 12GB, that’s even better for a smoother experience.

Tell Me More About It

Here’s the cool part. Fooocus learned from the best – it takes the offline, open-source, and free spirit of Stable Diffusion, and the simplicity of Midjourney. This means no excessive manual tweaking; just focus on what you want to create. It’s got all these neat optimizations and quality improvements built in, so you don’t have to worry about any technical stuff. Just think of it as a fun conversation between you and the computer to create something new and cool.

Installing Fooocus

Getting Fooocus up and running is a breeze. Here’s how you do it:

  1. Download the standalone installer from the GitHub repository. You will find it in the Download section.
  2. Unzip it and run the “run.bat” file, the Terminal will appear.
  3. When you first launch Fooocus, it automatically downloads the default models to a specific folder, so it might need some time to complete.
  4. Once that’s done, just open Fooocus in your browser using the local address provided, but it should do it automatically for you.

fooocus-ui-clean
Fooocus UI

That’s it, now you are ready to start generating images.


Generating Images

As I mentioned, there are already many optimizations behind the scenes, so you should expect great results by just typing a prompt and hitting generate.

Text to image

fooocus-ui-text-to-image

These results are great, and we didn’t use any custom model, LORA or any type of optimization, except for the ones made by Fooocus.

Image to Image

We can also use the image to image functionality by checking Input Image. This way, you will be able to generate variations based on an image that you load.

fooocus-image-to-image

I modified the prompt and got these results:

In this section you can also Upscale your image or create variations with a different level of strength. Keep in mind that upscaling takes a lot of resources, so it might not be as fast as a simple image generation.

Advanced Settings

We just saw how powerful Fooocus can be out of the box, but sometimes we might still want to be able to customize more parameters or choose different styles and models. In the advanced tab, you will find many interesting settings, like a set of predefined art styles, the possibility to use LORAs and select different custom models.

Settings Panel

In this section, you can handle parameters such as aspect rations, number of images to generate and performance. The performance session is an interesting one: you can prioritize speed over quality, by letting the model to run more steps if you prioritize quality. The latest addition is Extreme Speed, which uses LCM models to provide the best performances, with a slight drop in quality.

fooocus-fast-lcm
Six images generated with LCM at incredible speed

Style Panel

This is, in my opinion, the holy grail of Fooocus. A set of predefined styles to generate almost any kind of image with a simple selection. You can also select more than one style, to create unique combinations and to obtain exactly what you want. You will notice that some styles are preselected, like Fooocus V2, Enhance and Sharp – this is part of the premade optimization that are part of the UI, but it tends to generate realistic images with a similar style every time. So let’s select some different styles. for example Watercolor and SAI Anime.

A more recent update will show you some previews for each style, an amazing feature in my opinion. Just hoover the cursor on a style and you will see a small image of a cat in that style. Really handy!

Model Panel

Here you can choose different base models and LORAs, especially if the styles of the previous panel are still not enough for your needs. You can download models and lora on civit.ai or from other sources, you just need to put them in the proper folder: Fooocus_win64_2-1-25\Fooocus\checkpoints for models, and Fooocus_win64_2-1-25\Fooocus\loras for LORAs. As usual, you can adjust the strength of your LORAs using the sliders.

Advanced Panel

This is the latest panel available and you can adjust parameters about sharpness and guidance scale. This latest one, also known as CFG Scale, determines how strictly the model will follow the instructions of your prompts: lower values means that the model has more freedom while generating the image.

Conclusion

In summary, Fooocus stands out as an exceptionally user-friendly and efficient tool for anyone looking to start into the complex world of AI-generated art. It serves as the perfect starting point for beginners, offering an accessible UI that guides you through the power of Stable Diffusion XL.

3 thoughts on “Fooocus: The Easiest Way To Run Stable Diffusion With Impressive Results – Guide

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.