Krita AI Diffusion is a popular plugin that brings generative AI image creation directly into Krita, the open-source painting application. Instead of juggling a separate web UI, artists generate, inpaint and refine images on the canvas itself, powered by Stable Diffusion and Flux models through a ComfyUI backend. The plugin has about 10,800 stars on GitHub and is actively developed in 2026.
The plugin is designed around an artist’s workflow rather than a prompt box. Live painting gives near-real-time feedback as you sketch, inpainting and outpainting respect your existing layers and selections, and region-based controls let different parts of the canvas follow different prompts. It manages the ComfyUI server for you — install, generate, and iterate without touching node graphs — while advanced users can still connect their own ComfyUI setup and custom models, ControlNet-style guides and upscalers. If you lack a strong GPU, the project’s Interstice cloud service can run generation remotely and stream results back into Krita. Development is unusually transparent: the maintainer publishes detailed changelogs and workflow guides, and the plugin exposes advanced controls — generation strength per layer, seed control, specialized inpaint models and careful resolution handling — for artists who outgrow the defaults. It works on Windows and Linux with NVIDIA, AMD and Intel graphics, and recent versions added Flux model support alongside the Stable Diffusion family. For illustrators wary of cloud-only generators, it is one of the few AI tools that respects a traditional, layer-based painting process.
The plugin is free and open source under the GPL-3.0 license, and local generation costs nothing. The optional Interstice cloud service, for artists without a suitable GPU, is a paid subscription.
Krita AI Diffusion is for digital painters and illustrators who want AI generation woven into a real painting workflow, with full layer control instead of a standalone image generator.