Stable Diffusion Fooocus: A Guide For Beginners

If we talk about Stable Diffusion GUIs, the most popular names you’ll hear are Automatic1111 and ComfyUI. 

But there’s another great Stable Diffusion GUI that is not only fast but also very easy to use for beginners. 

Yes, I’m talking about Fooocus which is gaining popularity in the Stable Diffusion community for its simplistic UI and ease of use. 

In this Stable Diffusion Fooocus guide, I’ll walk you through this GUI starting from installation down to every feature available in it. 

By the end of this guide, you’ll fall in love with Fooocus and definitely give it a try. That being said, let’s get started. 

What is Fooocus? 

Fooocus is software to run Stable Diffusion offline with a simple and easy-to-use user interface. It’s open-source and completely free to use. 

It’s based on Gradio which means it has a very similar user interface to Automatic1111 since that’s also based on Gradio. 

Unlike Automatic1111, Fooocus provides a very simple way of generating images in Stable Diffusion quite similar to MidJourney. 

You don’t have to focus on aspects like sampling steps, CFG scale, models, etc when using Fooocus. It only puts your attention on the prompts and takes care of the rest. 

This gives you an experience similar to Dall-E or MidJourney where you enter prompts and generate images. 

Other than its easy interface, Fooocus also has a lot of technical optimizations that make it very performance efficient. 

The image generation times in Fooocus are much faster than Automatic1111 or ComfyUI. If you’re someone who has a lower-end GPU and is struggling to run Stable Diffusion, you’ll love using Fooocus. 

That’s because it’ll be able to generate images much faster. Moreover, you’ll be able to use SDXL models in Fooocus without any memory errors. 

How To Install Fooocus

Installing Fooocus is very simple and straightforward especially if you’re using Windows. However, you can also run it on Linus and Google Colab. 

Here are the steps to install Fooocus on Windows: 

Fooocus doesn’t have a complicated installation process for Windows. All you need to do is download Fooocus from the link below. 

Once downloaded, extract the zip file to any location you want and run the ‘run.bat’ file. That’s all. 

On your first startup of Fooocus, it’ll automatically download the checkpoint model which will take some time depending on your Internet connection. 

Once it’s downloaded, Fooocus will open as shown in the image below: 

If you want to use your checkpoint models in Fooocus, you can place them in the Fooocus/models/checkpoints folder directory. 

Besides the run.bat file in Fooocus, you’ll also find run_anime.bat and run_realistic.bat files. These are preset files that you can use if you want to generate images in that style. 

The run_anime.bat file uses the Bluepencil XL checkpoint model which is ideal for generating anime images in Stable Diffusion. 

The run_realistic.bat file uses the Realistic Stock Photo XL checkpoint model which is great at generating realistic images in Stable Diffusion. 

So, if you want to generate images in any of the above styles, you can choose to run Fooocus using these preset files instead of the run.bat file.

If you’re wondering whether you can change checkpoint models in Fooocus, the answer is yes. We’ll get into this later in the guide. 

How To Use Fooocus

As I mentioned above, using Fooocus is extremely easy and all you need to do is enter your prompts and generate the image. 

Let’s take a deeper look into how to use Fooocus and generate images using it.

Generating An Image

When open Fooocus, here’s how it will look like: 

As you can see, it’s just a blank screen with an option to write your prompt, a button to generate images, and two checkboxes for additional options. 

Let’s generate your first image in Fooocus. To do this, write a prompt for your image and click on the generate button. 

I’ll be using the following prompt for my image: 

1man, rooftop portraitCopy

Fooocus recommends you use very short and precise prompts as it takes care of the other aspects such as quality, looks, color, etc. You don’t need to write long and complicated prompts in Fooocus. 

Here are the output images generated by Fooocus: 

The images generated look absolutely gorgeous and we didn’t even have to scratch our heads writing a good prompt. That’s the power of Fooocus!

If you want to generate beautiful images in Stable Diffusion, Fooocus is the perfect companion as you don’t have to get your hands dirty with the nitty gritty of image generation. 

Just write your prompt and it’ll spit out an image that looks pretty. But if you want more from Fooocus, then you probably need to read the rest of the guide. 

Image Generation Settings

Fooocus only gives you the bare minimum at first to make image generation in Stable Diffusion as quick and simple as possible. 

But it does have a lot of features you can take advantage of if you want more control over the images you generate. 

Click on the Advanced checkbox visible right under the prompt field. 

This will display a sidebar on the right with more settings giving you more control over the output settings. 

There are four tabs namely: Settings, Style, Model, and Advanced. 

In the Settings tab, you’ll find a setting named Performance with three options: Speed, Quality, and Extreme Quality. 

Here’s what these three options do: 

• Speed: Uses 30 sampling steps to generate the image
• Quality: Uses 60 sampling steps to generate the image
• Extreme Quality: Uses LCM-LoRA to generate the image faster

Based on your GPU, you can choose any of the three performance options. I use the Speed option as it’s fast and still generates images that look amazing. Here are the differences between the three options: 

Below the Performance option is the Aspect Ratio setting with many choices you can pick for the image size. So, you don’t have to enter the image size manually and just pick from one of these popular image size presets. 

There’s also an option to choose the number of images you want to generate. By default, this number is set to 2 images but you can change it to any value you want. 

And lastly, there’s a text field where you can enter a negative prompt for your image if you want. I usually leave this empty since Fooocus is very good at generating images. But you can still write a short negative prompt if you want. 

There’s also a checkbox named Random which defines the seed of the image. If you uncheck this box, you’ll see an option to enter a seed number for your image. 

Unless you want to generate a specific image you previously generated or saw somewhere, you can set the seed to random. 

Style Settings

One of the best features of Fooocus is the style settings that let you generate images in so many styles presets. 

This setting can be found in the Style tab where you’ll find many options you can enable when generating an image. By default, three style options are already enabled. 

But there are many other diverse Stable Diffusion styles such as cinematic style, cyberpunk style, isometric style, ancient illustration style, etc. 

If you hover over the option, you’ll see a preview image showcasing what it looks like. This is very useful as you can see every style without trying it. 

Prompt Used:

a white rabbitCopy

Here are some examples of styles available in Fooocus:

Model Settings

The next tab is the Model settings where you can select the checkpoint model as well as apply up to 5 LoRA models. 

Since Fooocus supports SDXL, it also gives you the option to choose a refiner model in case you’re using SDXL models. 

I really love the fact that applying LoRA models is so simple in Fooocus. You just need to select the model from the dropdown and set the model weight. 

Besides the Model tab, there’s also a tab named Advanced where you can set the Guidance Scale (CFG) and the image sharpness. 

Image Upscaling

Fooocus also has another checkbox under the prompt field named Input Image that lets you input an image and use it for various purposes. 

When you enable the Input Image checkbox, it’ll open a new menu with 4 tabs. The first tab is the image upscale tab where you can upload your image and upscale it. 

Think of it as the Extras tab in Automatic1111 where you upscale an image. But unlike Automatic1111, all you need to do in Fooocus is to upload your image and choose your upscale variation. 

You can set the upscaling to be subtle, strong, 1.5x upscale, 2x upscale, or fast 2x upscale. You don’t have to worry about which upscaler to use or the upscaled image size or aspect ratio. 

Fooocus again simplifies this process and makes upscaling quick and easy. Here is a comparison of upscaling done in Fooocus. 

The variations options Subtle and Strong not only upscale the image but also change the image. This is similar to the option in MidJourney where you can generate variants of an image. 

Here is a comparison of variations done on an image in Fooocus: 

The difference between the variations isn’t much so I didn’t find this feature very useful. You’re much better off increasing the number of images generated if you want more variations. 

ImagePrompt

Fooocus lets you use an image as a prompt without the need to install extensions, unlike Automatci1111. 

To use it, switch to the ImagePrompt and upload your image to any of the four image slots. 

Now, click on the Advanced checkbox which opens up several options you can do with your image. There are several radio buttons that let you choose what you want to do with your image. 

ImagePrompt

The first option is ImagePrompt which lets you generate an image based on your input image. You can set the prompt weight and stop the image prompt control at certain sampling steps.

You can increase either value to increase the effect of the image prompt. 

PyraCanny

With PyraCanny, you can copy the composition or poses of humans or characters in your input image. PyraCanny is an edge-control method that detects edges in an image. 

It works similarly to the ControlNet canny method. 

CPDS

CPDS works just like the depth model of ControlNet and copies the 3D composition or depth of the input image. 

You can use it to copy the depth from an input image to copy composition details. 

FaceSwap

With the FaceSwap feature of Fooocus, you can swap faces easily and generate consistent faces or characters. 

It works like the IP-Adapter or the ReActor extension available in Automatic1111. 

Fooocus has four image slots allowing you to use these features in combination with one other. You can copy the composition of an image and apply face swapping to it. All without installing multiple extensions. 

Inpainting & Outpainting in Fooocus 

Fooocus also has features for inpainting and outpainting just like in Automatci1111 but more simply and straightforwardly. 

Switch to the ‘Inpaint or Outpaint’ tab and upload your input image. 

Then, you just draw a mask over the area you want to change write your prompt, and hit Generate. 

There are additional options for inpainting where you can specify the method to either improve details (such as hands, eyes, etc) or modify the content of the image. 

Prompt Used: 

wearing sunglassesCopy

Here are the results: 

With Outpainting, you can extend your image in any direction you want. To use outpainting in Fooocus, upload your image and set the method to ‘Inpaint or Outpaint (default)’. 

Now, set the Outpaint direction to any of the desired options you want. 

You don’t have to enter any prompt during outpainting. However, you can still write a prompt if you wish to. 

Describe

Lastly, Fooocus has the option to describe any image to guess its prompt. This feature is very similar to the Interrogate CLIP feature of Automatic1111. 

Just upload your image in the Describe tab and click on the ‘Describe this image into prompt’ button. 

The prompt it guesses will appear in the prompt field. 

Fooocus vs Automatic1111 vs ComfyUI

So, you’re now familiar with all the features available in Fooocus. But how does it actually compare to the likes of Automatic1111 or ComfyUI? 

Here’s a detailed comparison table outlining the differences and similarities between Fooocus, Automatic1111, and ComfyUI. 

Pros and Cons of Fooocus 

Here are the pros and cons of Fooocus:

Pros

• Very easy to use and get started
• Perfect for beginners who just want to generate images
• Most features are simplified 
• Generates high-quality images without much configuration
• High performance that is ideal for lower-end GPUs 

Cons

• Cannot install extensions 
• Is very limited for any complex workflow 
• Inpainting is very limited 
• Not ideal for complex workflows 

FAQs

Here are some frequently asked questions about Fooocus: 

Conclusion

If you want straightforward software to run Stable Diffusion, then Fooocus is the best option out there. 

Yes, it’ll feel somewhat limited when compared to heavyweights like Automatic1111 and ComfyUI. But sometimes you just want to generate beautiful images quickly without the hassle of changing the CFG scale, samplers, and whatnot. 

And in those cases, Fooocus comes in handy. It gives you an experience of using Stable Diffusion like Dall-E or MidJourney where you just enter a prompt and generate your image. 

If you have any questions or doubts about Fooocus, feel free to ask your questions in the comments below.