Fooocus

Fooocus is a free and open-source image generation program built on Stable Diffusion XL, created by Lvmin Zhang, the author of ControlNet. Its design philosophy combines the offline and open-source nature of Stable Diffusion with the ease of use and default quality of Midjourney, allowing users to focus on prompts and creative ideas rather than technical parameters.

The software features a simple installation process requiring fewer than three mouse clicks from download to first image generation. It includes GPT-2 based prompt enhancement that automatically adds detail to user prompts, over 200 built-in style presets, and carefully tuned default settings that produce high-quality results without manual configuration. Advanced users can access a full settings panel for fine-grained control over sampling methods, guidance scale, and other parameters.

Fooocus implements several technical innovations including advanced k-diffusion sampling for seamless base-refiner model swapping, negative ADM guidance to prevent plastic or overly smooth outputs, and Self-Attention Guidance (SAG) for consistent image quality. The software supports inpainting, outpainting, LoRA models, and multiple Stable Diffusion model variants. It can run locally on consumer GPUs or in Docker containers on cloud services.

The project is currently in long-term support mode with bug fixes only, as the creator has moved focus to WebUI Forge. For users wanting newer model architectures like Flux, the team recommends WebUI Forge or ComfyUI. Despite this, Fooocus remains one of the most popular entry points for local AI image generation because nothing else is simpler to set up.