Fooocus: creating images locally with open and easy tools
Image generation is the missing half of your complete Ollama guide: while language models take care of text, a whole family of local tools can now make images. The trouble is that the best-known tool — ComfyUI — is also one of the most intimidating to install and learn. The good news: it is not the only path.
This guide reviews the free tools that run entirely on your machine, compares how easy they are to install and clearly picks which one to choose depending on your profile. As a bonus, we detail the complete installation of our recommendation.
Why create images on your machine
Before comparing, let's recap the point: your text and your images never leave your computer (privacy), every extra image is free once the hardware is bought (cost) and everything works offline. It's the same argument that drew you to local AI for text.
The criteria that matter
To compare tools honestly we look at three things:
- License: is the software truly free (open source)? Be careful: some tools are free but the models they ship have restrictive licenses.
- Ease of installation: does it need a terminal, commands, tinkering? Can you pick the easy path with an installer?
- Ease of use: once launched, does it produce a nice image from simple text or do you have to tune dozens of settings?
The comparison
| Tool | Software license | Installation | Use | Who it's for |
|---|---|---|---|---|
| Fooocus | Free (GPL-3.0) | Simple installer, 1 command | Immediate, Midjourney-style | Beginners, everyone |
| Stable Diffusion WebUI (Forge) | Free (AGPL/GPL) | Install script, more settings | Rich interface, more control | Intermediate and curious users |
| InvokeAI | Free (GPL-3.0) | Modern graphical installer | Good balance | Intermediate users |
| ComfyUI | Free (GPL-3.0) | More demanding (block graphs) | Very powerful but complex | Experts |
The model trap
All these tools let you choose the model that makes the images. This is where the nuance hides:
- Stable Diffusion 1.5 and SDXL: open weights and permissive licenses — the free models par excellence.
- Stable Diffusion 3 and Flux: the weights are accessible but the license is restrictive (non-commercial use for some, fine-tuning limits for others).
The software is free, but you must check the license of the model you use. This distinction is detailed in our article open source vs open weight.
Our recommendation: Fooocus
If you want the fastest path to a nice image without knowing anything about AI, Fooocus is the obvious choice.
Why:
- Built to start easy: it is designed to feel like Midjourney — a single box where you type and get a good-looking image.
- Well tuned by default: the fine settings — number of passes, model choice, refiner — are picked intelligently for you. It is "optimized by default".
- Free: GPL-3.0 license, built on Stable Diffusion.
- Easy to install: on every platform the install is a few lines.
Installing Fooocus
Prerequisites
- A graphics card is strongly recommended (NVIDIA, AMD or Apple Silicon). Without one a CPU mode exists but is slow.
- 8 GB of RAM minimum, 16 GB recommended.
- Disk space: Fooocus downloads a model at first launch (several gigabytes). Make room.
- Python 3.8+ on your machine.
On Windows and macOS
Fooocus ships a graphical installer you download from its official site. You install it like any software, launch it and the interface opens in your browser. This is truly the simplest path: no command line at all.
The official page is github.com/lllyasviel/Fooocus (the Releases tab for ready-made installers).
The manual path (the same everywhere, including GNU/Linux)
For full control or on systems without a graphical installer, everything runs from the command line:
# Get the software
git clone https://github.com/lllyasviel/Fooocus.git
cd Fooocus
# Create a clean environment
python3 -m venv venv
./venv/bin/pip install -r requirements_versions.txt
# Launch
./venv/bin/python entry_with_update.py
The first launch downloads a model (weighing several gigabytes), then the interface opens automatically in your browser. That's it.
Your first image
- Open the interface in your browser.
- Type your text in the box.
- Click Generate.
That's it. Images appear in a few seconds to a few dozen seconds depending on your hardware. Advanced options (size, style, number of images) stay available in the tabs if you want to dig in, but you don't need them to start.
What hardware you actually need
| Hardware tier | What runs comfortably |
|---|---|
| CPU only | Possible but slow, minutes per image |
| 4-6 GB video memory | Simple models, modest images |
| 8-12 GB | Comfortable images, common models |
| 16 GB and above | Total comfort, large models |
| Apple Silicon (shared memory) | Very good with 16 GB and more |
If Fooocus is not enough
Fooocus is deliberately simple. If you get hooked and want more control — fine settings, LoRA, control — move on to Stable Diffusion Forge or InvokeAI, which keep the install reasonable while offering many more levers. This guide is a good starting point, not a destination.
Conclusion
To create images locally, easily and with free tools, Fooocus is the best entry point: installable in a few minutes on Windows, macOS and GNU/Linux, usable with zero technical knowledge and completely free. The software is and if you stick to open-license models like SD 1.5 or SDXL, your whole stack is free end to end.