Fooocus on Windows: setup, inpainting and outpainting

This tutorial walks through installing Fooocus on Windows for local AI image generation. The speaker says 6 to 8 GB of NVIDIA VRAM is enough and lists CUDA Toolkit 11.8 to 12.8, Visual Studio, FFmpeg and Git for Windows as prerequisites. Those requirements are the tutorial's stated setup, rather than a guarantee for every configuration.

The installation starts with cloning the Fooocus repository and opening its folder in a command window. The speaker creates and activates a Python 3.10 virtual environment, upgrades setuptools, wheel and pip, then installs PyTorch with a CUDA version matching the installed toolkit. The walkthrough also adds torchaudio and edits the requirements file to replace ONNX Runtime with ONNX Runtime GPU at a version the speaker describes as compatible with PyTorch.

After installing the requirements, the tutorial includes a Hugging Face Hub installation step. The transcript refers to a specific version but does not state its number. Launching Fooocus then downloads models and opens the interface. A run.bat file provides a shortcut for later launches.

The final section covers advanced settings such as model style and resolution, followed by an image generation example. The speaker describes text-to-image, image-to-image, inpainting for replacing parts of an image, and outpainting for extending its borders. The editing features receive a brief introduction rather than a detailed demonstration.