What is Fooocus?

While StableDiffusionWebUI is known for its robust functionalities, many individuals find the simplicity of prompt-based interaction in Midjourney appealing. Drawing inspiration from Stable Diffusion and its extensions, like ControlNet, the developer behind ControlNet, lllyasviel, has launched Fooocus.

Simplified Interface

Fooocus is designed to streamline the user experience, allowing creators to concentrate solely on their prompts without the distraction of numerous settings. This reimagined version of Stable Diffusion, influenced by the prompt-centric methods of ComfyUI and Midjourney, automates the handling of various settings, aiming to deliver high-quality image generation with ease.

The primary interface of Fooocus, similar to StableDiffusionWebUI, utilizes Gradio. However, it features a minimalistic design, comprising just a prompt input field and a Generate button, ensuring a clutter-free experience.

Reduced Memory Requirements

Another significant advantage of Fooocus is its low VRAM consumption for generating SDXL images. The software claims that it can operate effectively with as little as 6GB of VRAM. This marks a considerable improvement, as the previous SD 1.5 version demanded significantly more VRAM for image generation. As a result, users with limited VRAM can now seamlessly create SDXL images.

This powerful and simplified functionality is a welcome development for many users. Are you excited to give it a try?

Installation of Fooocus

Windows Version

1. Download the zip file from GitHub:

2. Unzip the file and execute run.bat to install.

   During the initial run, the software will automatically download large files exceeding 6GB each, which may take some time.

MacOS / Linux Version

1. If you haven't already, refer to the installation guide for StableDiffusionWebUI and Python setup. Skip these steps if Python is already installed.

2. Use Git to download the Fooocus files:

3. Set up a virtual environment (venv):

   python -m venv ./venv

4. Activate the virtual environment:

   source ./venv/bin/activate

5. Install the necessary Python packages:

   pip install -r requirement_versions.txt

6. Launch Fooocus:

   python launch.py

   Similar to the Windows installation, this process will also download large files during the first run.

Basic Usage

Utilizing Fooocus is incredibly straightforward. Simply enter your prompt in the designated field and click Generate. By default, it produces two images at a resolution of 1152 x 896 with a cinematic aesthetic.

Advanced Settings

You will notice an Advanced button located below the prompt input field, leading to further customization options.

• Performance: The default Speed option generates images in 30 steps, while the Quality setting takes longer, employing 60 steps for improved results.
• Aspect Ratios: The default resolution is 1152 x 896.
• Image Number: You can generate up to 32 images at once.
• Random Seed: The default value of -1 generates a random seed, but you can specify a custom seed number.
• Negative Prompt: This allows you to indicate elements or concepts to be excluded from the generated images.

Styles

The styles section allows you to select from over 100 distinct artistic styles, with cinematic-default being the initial setting. While style names offer a glimpse into their essence, visual references can be more enlightening. To facilitate this, I generated an image for each style using the prompt "1girl" at a resolution of 1024 x 1024, providing a practical comparison tool. (Creating these reference images required more time than writing this post, especially since the official source lacks such visual aids.)

Advanced Options

In this section, you can further refine model settings and LoRAs.

• SDXL Base Model: Default automatically downloads sd_xl_base_1.0_0.9vae.
• SDXL Refiner: Default automatically downloads sd_xl_refiner_1.0_0.9vae.
• LoRAs: You can select up to five LoRAs at once, along with their respective weights. By default, the software employs the sd_xl_offset_example LoRA, enhancing the output quality.

Conclusion

Though still in its early stages, Fooocus has already generated significant enthusiasm. Its user-friendly interface and efficient resource management make it an accessible tool for those who find Stable Diffusion initially challenging. We look forward to seeing how Fooocus evolves into a powerful yet intuitive software, inviting more individuals to explore AI-assisted art creation.

A tutorial comparing Automatic1111 and Fooocus, showcasing how to use Stable Diffusion UI effectively.

An introduction to Fooocus, the easiest user interface for Stable Diffusion XL, demonstrating its unique features and capabilities.

Connect with us on LinkedIn for the latest updates and insights in the realm of AI. Let's shape the future of artificial intelligence together!