Krita AI Diffusion

Generative fill inside a professional painting app is what Krita AI Diffusion delivers: select a region on the canvas, describe what belongs there, and the plugin inpaints, outpaints, refines, or upscales toward 4K and 8K without leaving Krita. A live painting mode reinterprets brush strokes as finished art in real time, region prompts assign separate text to separate areas, and ControlNet scribble, line art, depth, and pose conditioning plus IP-Adapter reference images give illustrators fine control. Under the hood the plugin drives a ComfyUI server, either installed automatically on the local machine or reached through the developer's paid interstice.cloud service, and it supports Stable Diffusion 1.5, SDXL, Illustrious, and Flux model families. Local generation is recommended with a GPU of about 6 GB VRAM and runs on NVIDIA CUDA, AMD ROCm, Intel XPU, Apple Silicon, or slowly on CPU. Installation goes through Krita's Python plugin manager on Krita 5.2 or newer; the project is GPL-3.0, actively developed, and documented at docs.interstice.cloud.