Krita AI Diffusion: local setup and photo editing course

This course explains how to set up Krita AI Diffusion and use local AI for image generation and photo editing. The speaker demonstrates Windows Portable, imports the Python plugin, and installs a managed local server. Krita downloads for macOS and Linux are mentioned, but the setup walkthrough uses Windows. Plugin compatibility needs checking against the installed Krita version.

The setup includes Flux 2 with 4 billion parameters and a manual installation of 9B weights with a Qwen 3 8B GGUF text encoder. The speaker prefers 9B for generation and considers 4B sufficient for object editing. For comfortable use of the larger model, he recommends at least 12 GB of video memory; weaker hardware may run it slowly. He says 16 GB RAM can suffice with heavy applications closed and recommends 32 GB for professional design work.

Editing examples cover removing objects and text, adding furniture, replacing backgrounds, and building a product composition with reference images. The course also covers colorization, damaged-photo restoration, background removal, upscaling, canvas expansion, and turning sketches into photographic images. Generated edits remain on separate layers for further adjustment.

The examples expose limits as well as useful results: additions can be unpredictable, generation can damage lettering, and restoration can alter faces. The speaker demonstrates cleanup with brushes, masks, opacity adjustments, and lower refinement strength.