Fooocus
A local AI image generator using Stable Diffusion XL. Runs offline on Windows, Linux and Apple Silicon Macs, with image editing and upscaling.
Fooocus is a free, open-source AI image generator for people who want to create images on their own computer without spending much time tuning settings. It uses Stable Diffusion XL and automatically expands prompts with a local GPT-2 engine. Generation works offline once the required models are downloaded, so prompts and images can stay on your machine.
Text prompts are only part of it. You can use an existing image as a reference, generate subtle or stronger variations, and upscale results. Inpainting replaces selected areas, while outpainting extends an image beyond its original borders. Fooocus includes its own models and methods for inpainting and image prompts, plus face swapping through InsightFace and a Describe feature that turns an image into text.
Presets cover general images, realistic photography and anime. It also accepts SDXL models from Civitai. Style choices, negative prompts and prompt weighting give you more control when the automatic defaults don't produce what you want.
The browser interface runs locally on Windows or Linux, with Docker support and Apple Silicon acceleration on macOS. NVIDIA RTX systems need at least 4GB of VRAM and 8GB of system RAM; CPU-only generation needs 32GB of RAM and is much slower. Mac support has had limited testing. Google Colab is an alternative, but generation there runs on cloud hardware.
Fooocus uses the GPL-3.0 license and remains focused on SDXL. The project is in limited long-term support, with maintenance restricted to bug fixes.
