How to Use Fooocus in 2026: Complete Guide and Pro Tricks

Fooocus is a free tool that brings together the best features of Stable Diffusion and Midjourney. It's designed to be open source, work offline, and be easy to use. With Fooocus, you can create high-quality images without spending hours adjusting settings.

In this guide, we'll walk you through how to use Fooocus and share some helpful tricks to get the most out of this tool. If you want the long-form reference on the project, presets, LoRAs, and how it compares to ComfyUI and Forge, read our Fooocus 2026 complete guide. This page is the hands-on version: install, first image, the settings that matter, and the tricks that save time.

What Fooocus Is

Fooocus is an open-source Stable Diffusion frontend, created by lllyasviel (the researcher behind ControlNet), that runs SDXL models on your own GPU behind a single prompt box. It hides samplers, schedulers, and refiner settings behind presets, so a beginner gets a good image on the first try, while an Advanced checkbox exposes the full controls when you need them. The source code and releases live at github.com/lllyasviel/Fooocus.

Hardware You Need

Fooocus runs on a 4 GB NVIDIA GPU with its built-in memory optimizations, and 8 GB is the comfortable recommendation. More VRAM lets you stack LoRAs and use the refiner without swapping.

You also need roughly 8 GB of free disk space for the default SDXL checkpoint and refiner, plus space for any extra models you download.

Getting Started with Fooocus

Installation

Windows (easiest path):

1. Download the Fooocus zip file from the official GitHub releases page.
2. Extract the contents to your preferred folder.
3. Run the 'run.bat' file to start Fooocus.

The package also ships 'run_realistic.bat' and 'run_anime.bat', which launch the same app with the Realistic or Anime preset already selected. Pick the one that matches what you want to make most often.

Note: The first time you run it, Fooocus will download necessary models. This might take a few minutes depending on your internet speed.

Linux and macOS:

1. Install Python 3.10 or 3.11 (newer Python versions have dependency issues with some packages).
2. Clone the repository with git clone https://github.com/lllyasviel/Fooocus.git.
3. Create a virtual environment, install the requirements with pip install -r requirements_versions.txt, and launch with python entry_with_update.py.

On Apple Silicon the app runs through MPS instead of CUDA. It works, but expect each image to take several times longer than on a mid-range NVIDIA card.

Cloud (no GPU):

If you don't have a compatible GPU, the official Colab notebook runs Fooocus in your browser: fooocus_colab.ipynb. Free Colab sessions are limited, so save your outputs regularly. Per-second GPU rentals like RunPod are the next step up when you need longer sessions.

Basic Usage

1. Open Fooocus
2. Type your image description in the prompt box
3. Click 'Generate'

It's that simple to start creating images!

Your First Good Image

A short, concrete prompt beats a long one in Fooocus, because the preset already appends quality terms for you. A working prompt has a subject, a few style words, and nothing else:

portrait of a woman in a rain jacket, city street at night, neon reflections, 85mm lens

Leave the negative prompt empty on your first run. Fooocus applies its own prompt expansion (the "Fooocus V2" style) so you don't need "masterpiece, best quality" boilerplate. If a word doesn't visibly change the image, remove it. For a library of tested prompts by genre, see our Fooocus image prompts guide.

The Settings That Matter

Tick the 'Advanced' checkbox under the prompt to reveal the settings panel. Five controls account for almost every difference in output.

Two more live under the Advanced tab: Guidance Scale (how literally the prompt is followed; the default works for most subjects, raise it slightly if the image ignores your prompt) and Image Sharpness (raise for crisp product shots, lower for soft portraits). Change one at a time with a fixed seed so you can see what each does.

Prompt Weighting Syntax

Fooocus uses the standard SDXL weight syntax, so prompts from other Stable Diffusion tools carry over unchanged:

• (keyword) adds emphasis
• (keyword:1.4) sets an explicit weight, here 40 percent stronger
• [keyword] reduces emphasis

Example: portrait of a woman, (cinematic lighting:1.4), (sharp focus:1.2), [oversaturated]. Weights above roughly 1.5 tend to distort the image, so push one or two terms, not the whole prompt.

Fooocus Tricks for Better Results

1. Use the Style Menu

Fooocus comes with preset styles that can dramatically change your output. Here's how to use them:

1. Check the 'Advanced' box at the bottom of the interface
2. Look for the 'Style' dropdown menu
3. Experiment with different styles like 'Cinematic', 'Anime', or 'Photographic'

Tip: You can combine multiple styles for unique effects.

2. Adjust Performance Settings

Balance speed and quality based on your needs:

• 'Speed': Good for quick drafts
• 'Quality': Best for final images
• 'Extreme Speed': Use when you need results fast, but expect lower quality

3. Use Image Prompts

Guide Fooocus with reference images:

1. Check 'Input Image'
2. Select 'Image Prompt' tab
3. Upload your reference image
4. Adjust 'Stop At' and 'Weight' sliders to control the influence

4. Try Inpainting and Outpainting

Modify specific parts of an image or extend it:

1. Upload an image
2. Select 'Inpaint or Outpaint'
3. Use the brush to mark areas you want to change
4. Add a text prompt to guide the changes

The Method dropdown (Improve Detail, Modify Content, Remove Object) decides what the mask does, and getting it wrong is the most common reason inpainting "does nothing". Our Fooocus inpainting tutorial covers methods, denoise strength, and seam fixes in depth.

5. Experiment with Aspect Ratios

Create images in various sizes:

1. Look for the 'Aspect Ratios' dropdown in the main interface
2. Choose from preset sizes or add custom ones in the config file

6. Upscale or Vary a Result You Like

The 'Upscale or Variation' tab under Input Image is the fastest way to iterate on a good image without re-rolling the seed. 'Vary (Subtle)' keeps the composition and changes small details, 'Vary (Strong)' reinterprets the image more freely, and the Upscale options enlarge the image while adding detail. Generate at the default size, pick the best candidate, then upscale only that one.

7. Describe an Image to Get Its Prompt

The 'Describe' tab takes an uploaded image and writes a prompt for it, with separate modes for photos and anime. It is the quickest way to reverse-engineer a look you want to reproduce, and pairs well with Image Prompt for composition control.

8. Fix the Seed Before You Tune

Untick 'Random' next to the seed and reuse the same number while you adjust styles, weights, or guidance. With the seed fixed, every change you see comes from the setting you touched, not from a new roll of the dice.

Advanced Fooocus Features

Custom Model Integration

Use your favorite Stable Diffusion models:

1. Place your model files in the 'models/checkpoints' folder
2. Select your model from the 'Model' dropdown in the advanced settings

Any SDXL checkpoint works, and the same files are interchangeable with ComfyUI. See our picks for the best SDXL models for realism, anime, and all-purpose checkpoints.

LoRAs

LoRAs are small add-on files that teach the base model a style, character, or concept. Drop them into 'models/loras', then pick them in the LoRA slots under the Model tab and set a weight, usually between 0.6 and 0.8 to start. Fooocus supports several LoRAs at once, but stack them only when each one pulls a different dimension. The full walkthrough is in How to Add and Use LoRAs in Fooocus.

Prompt Expansion

Fooocus automatically expands your prompts. To see the expanded version:

1. Generate an image
2. Check the 'log.html' file in the output folder

The log also records the seed, styles, and every setting for each image, which makes it the easiest way to reproduce a result weeks later.

Multi-Image Prompts

Combine multiple reference images:

1. Upload multiple images to the Image Prompt slots
2. Adjust individual weights and stop points for each

Troubleshooting Fooocus

• If images aren't generating, check your GPU compatibility and drivers
• For unexpected results, try clearing your prompt and starting fresh
• If the interface is sluggish, close other resource-intensive programs

Fooocus vs ComfyUI and Other Frontends

Fooocus is the right choice when you want good SDXL images with the least setup and no node graphs. ComfyUI wins once you need multi-stage pipelines, the newest model families, or fine control over every sampling step; Forge sits in between with an Automatic1111-style interface. Our Fooocus vs ComfyUI vs Automatic1111 comparison lays out the trade-offs, and the ComfyUI 2026 guide is the place to start if you outgrow Fooocus.

By mastering these techniques, you'll be able to create impressive images with Fooocus efficiently. Remember, the key to great results is experimentation. Don't be afraid to try new combinations of settings and prompts!

FAQ

Is Fooocus free?

Yes. Fooocus is open source under the GPLv3 license and costs nothing to download and run. The Stable Diffusion models it loads have their own licenses, so check the model's terms before commercial use.

What is the easiest way to install Fooocus?

On Windows, download the release zip from GitHub, extract it, and double-click 'run.bat'. It installs its own Python environment and downloads the default models on first launch, so there is nothing else to configure.

How much VRAM does Fooocus need?

A 4 GB NVIDIA GPU is the minimum thanks to Fooocus's built-in memory optimizations, and 8 GB is recommended. With 12 GB or more you can use the refiner and stack several LoRAs comfortably.

Does Fooocus work on Mac?

Yes, on Apple Silicon Macs through MPS. Generation is much slower than on an NVIDIA card, so it suits testing and small batches rather than production work.

Can Fooocus run FLUX or SD 3.5 models?

Not in the default install. Fooocus is built around SDXL, and newer model families are better served by ComfyUI. Any SDXL checkpoint or LoRA, however, drops straight into Fooocus.

Where are my generated images saved?

In the 'outputs' folder inside your Fooocus directory, organized by date, alongside a 'log.html' file that records the prompt, expanded prompt, seed, and settings for every image.