Fooocus
An offline, prompt-first image generator built on Stable Diffusion XL, now in bug-fix-only support.
Type a prompt and generate; no parameter tuning. The project is in bug-fix-only support.
Fooocus is an offline, open-source and free image generator on Stable Diffusion XL, designed so you focus on prompts rather than settings. It suits people who want a Midjourney-style flow on their own PC and are happy with SDXL only.
What it is
You unzip the Windows package and run run.bat, or install on Linux and macOS with Conda or a venv. The first run downloads the SDXL models automatically. It includes an offline GPT-2 based prompt processor and its own inpaint and outpaint algorithms.
Who it's for
It suits beginners with a 4 GB or larger Nvidia card who want simple SDXL images. It does not suit anyone who wants Flux or newer models; the README says no newer architectures are planned.
Key features
Prompt-first workflow
Aims to need little prompt engineering or tuning.
Offline prompt engine
A GPT-2 based processor expands prompts locally.
Inpaint and outpaint
Its own algorithm and models handle edits and extending images.
Prompt weights
Uses the A1111 reweighting syntax such as (happy:1.5).
Presets
Default, Anime and Realistic presets are documented.
Low minimum specs
4 GB Nvidia VRAM and 8 GB RAM minimum.
Up and running in 3 steps
Our honest verdict
What's great
- Windows install is unzip and run run.bat, with SDXL models downloaded on first launch.
- GPL-3.0 code with the README listing hardware minimums per GPU family.
- Runs on as little as 4 GB of Nvidia VRAM with 8 GB RAM and system swap.
- Its own inpaint and outpaint models give edit tools without extra plugins.
Where it falls short
- The README declares limited long-term support with bug fixes only, so no new features or models.
- It is built only on Stable Diffusion XL, so Flux and newer architectures are out of scope.
- CPU-only use needs 32 GB RAM and the README puts it at about 17 times slower than an RTX 3000-series card.
- Memory-allocation errors may need at least 40 GB free space on each drive, per the README.
Pricing & license
Free under GPL-3.0. The only cost is the Nvidia, AMD or Apple hardware you run it on; the project shows no paid hosted plan.
Open remote AI-training roles
768 open remote AI-training roles, updated November 14, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Is Fooocus still being developed?
The README says it is in limited long-term support with bug fixes only and no current plans to adopt newer model architectures.
What GPU does Fooocus need?
The README gives a 4 GB Nvidia VRAM minimum with 8 GB RAM and swap on RTX 2000 to 4000 cards, and 8 GB for AMD.
Can I use Fooocus images commercially?
The code is GPL-3.0. The SDXL models it downloads have their own licences, so check them before commercial use.
