Krita AI Diffusion on Linux: compatible versions, installation and common problems
Install Krita AI Diffusion on Linux with the right plugin version. Understand AppImage, local backends, GPU requirements and common setup problems.
Krita AI Diffusion brings generative image editing into Krita, including filling a selected area and extending an image beyond its edges. On Linux, getting started means matching three things: your Krita version, the plugin release and the backend that will generate images.
The plugin is a separate project from Krita. Its installation requirements and support channels come from the plugin developers. Our Krita AI Diffusion download page covers its broader features; this guide focuses on the documented setup and the problems to check first.
Choose a plugin release that matches Krita
1.53.0 release targets Krita 5, with Krita 5.3.2.1 recommended. The 1.54.0 prerelease targets Krita 6, recommends Krita 6.0.4 and explicitly drops compatibility with Krita 5.x.
| Your application | Release to investigate | Important distinction |
|---|---|---|
| Krita 5 | 1.53.0 stable release | Published for Krita 5 |
| Krita 6 | 1.54.0-pre | Prerelease, not the stable default |
AppImage or Flatpak?
For a managed local backend, it's recommended to use Krita's AppImage, as some packages cannot be installed through the Flatpak build.
Choose the backend before downloading models:
| Backend | Where generation runs | Who handles setup |
|---|---|---|
| Local Managed Server | Your computer | The plugin installs the server |
| Custom Server | Local or remote ComfyUI | You maintain the server and dependencies |
| Online Service | Cloud GPUs | You need a service account |
Install the plugin ZIP
The recommended sequence is:
- Download the packaged ZIP.
- Open Tools > Scripts > Import Python Plugin from File in Krita.
- Select the ZIP, enable the plugin and restart Krita.
- Open a document.
- Enable AI Image Generation under Settings > Dockers.
Configure the backend from that panel. With a managed server, select only the workloads you need initially.
Connecting an existing ComfyUI installation
A ComfyUI server that works in a browser is not automatically ready for the Krita plugin.
The project's custom-server guide lists required node extensions and model files. The plugin checks these when connecting. Missing dependencies can therefore prevent a connection even when the server itself starts normally.
Reusing a server can make sense when you already maintain one and understand its model paths. For a first installation, the managed option reduces the number of components you must assemble yourself.
Make a first selection fill
For a small first exercise, use a copy of an image, select one uncomplicated area and describe what you want there. This keeps the task easier to assess than replacing most of the canvas at once. The result will still depend on the model and settings.
Generated images initially appear as previews. Applying a result adds it as a layer; seeing a preview does not mean you have already committed it to the document. Lower generation strength also gives existing image content more influence during refinement.
Common problems to check first
| Symptom | Documented check |
|---|---|
| Plugin entry is greyed out | Confirm you installed the release package, not GitHub's source ZIP; inspect the entry's error tooltip |
| An update seems to have no effect | Restart Krita after replacing the plugin |
| Custom server will not connect | Check required extensions and models against the custom-server guide |
| Installation or generation fails | Open the logs from the plugin's connection settings and record the actual error |


No comment yet
Be the first to share your thoughts.