Short answer
For making many images from prompts across many models, HEISS UI is the quicker tool: 21 image families and video, each set up for you, one searchable gallery, and a phone studio. It adds nothing to your ComfyUI up front.
For working on one picture, with selections, inpainting, sketches and control layers, Krita AI Diffusion is the better tool. They can share one ComfyUI, so you can use both: HEISS UI to find the image, Krita to finish it.
Pick HEISS UI if
- You mostly write prompts, compare results and keep the good ones.
- You switch between many model families, including Krea 2, Qwen-Image and Chroma.
- You want video too: Wan 2.2, HunyuanVideo 1.5 and MiniMax H3, in beta.
- You’d rather not add node packs to your ComfyUI before you start.
- You want a gallery, a phone studio and a locked Hidden space.
Pick Krita AI Diffusion if
- You draw or paint, and want AI inside a real painting app.
- You inpaint and outpaint with selections, or turn sketches and line art into images.
- You use control layers, reference images, IP-Adapter and regions per layer.
- You like live painting, where the image follows your brush.
At a glance.
| Feature | HEISS UI | Krita AI Diffusion |
|---|---|---|
| Your ComfyUI | As it isnothing added up front | Needs four node packsor its managed install or paid cloud |
| Image models | 21 familieseach set up for you | Main familiesFlux 2, Z-Image, SD 1.5, SDXL, Anima |
| Video | Beta7 families | Not in its docs |
| Gallery | One for everythingsearch, stars, trash, old folders | Historyin the plugin |
| On a phone | Phone studiomade for one thumb | Desktop app |
| Private images | Hiddenencrypted | No |
| Custom workflows | Drop in the JSON | Yes |
| Upscale | One clickwith a compare slider | To 4k and beyondtiled |
| Interface | Prompt box and gallery in the browser | A docker panel in Krita’s canvas |
| Inpaint, outpaint | No | Yeswith selections |
| Control, regions | In imported workflows | YesControlNet, IP-Adapter, regions |
| Live painting | No | Yes |
| Licence | MIT | GPL-3.0 |
Checked 29 September 2026 against the plugin’s README, release notes and docs.interstice.cloud.
Painting or prompting.
Krita AI Diffusion’s goal, in its own README, is precision and control inside an image-editing workflow. You select an area and fill it, expand the canvas, refine at a chosen strength, guide the result with sketches, line art, depth or pose, and give each layer its own description. Live painting redraws as you paint. It’s the tool for people who already think in layers and brushes.
HEISS UI has no brushes. You write a prompt, previews stream in, and the gallery fills up. To change an image you give it a start image with a Change slider, or reference images for the edit models Flux.2 and Qwen-Image 2.1, and upscale the keepers with SeedVR2. If you need a precise fix in one corner of a picture, Krita is the better tool.
Same ComfyUI, two front ends.
Both send their work to ComfyUI. Krita’s plugin can install and manage a ComfyUI for you, connect to one you already run, or use its paid cloud service. HEISS UI always uses yours. If a ComfyUI is already running on this computer, both find it: the plugin tries to connect to a local server by itself, and HEISS UI checks ports 8188 and 8000.
Model coverage differs. The plugin’s model lists centre on SDXL, Flux 2, Z-Image and SD 1.5, with Anima added in 2026. HEISS UI runs 21 image families, from SD 1.5 to Krea 2, Chroma, Qwen-Image and Ideogram 4, plus video in beta.
What each needs installed.
Connected to your own ComfyUI, Krita AI Diffusion needs four custom node packs: ControlNet preprocessors, IP-Adapter, and Acly’s inpaint and tooling nodes, plus a set of models for the families you use. It doesn’t download those for a custom server; a download_models.py script does, and the managed install handles them for you.
HEISS UI adds nothing to ComfyUI up front. When a model needs a text encoder, a VAE or a node pack you don’t have, its setup lists it with a Download or Install button.
For a custom server, the plugin lists four required node packs and the models to add, and points to a script that downloads them.
The managed installer lets you pick workloads, packages of models, and reckons with 10 GB at least and often 50 GB or more.
Where Krita is the better pick.
Selections, inpainting, outpainting, sketch and line art to image, control layers, regions and live painting. For finishing a picture by hand, Krita is the tool.
Using both on one ComfyUI.
Start ComfyUI
The one you already use, or ComfyUI Desktop. Both front ends find it on this computer.
Set up the plugin’s needs
Install the four node packs and models from the plugin’s ComfyUI setup page, or run its
download_models.py.Open both
Generate and pick in HEISS UI, then open the image in Krita to paint, inpaint and finish. The files are in ComfyUI’s output folder.
Questions.
Does Krita AI Diffusion use ComfyUI?
Yes. It uses ComfyUI as its backend, either an install the plugin manages, your own ComfyUI, or its paid cloud service.
Can Krita AI Diffusion and HEISS UI share one ComfyUI?
Yes. Both connect to a ComfyUI running on this computer. The plugin needs its four node packs and models installed there first.
Does HEISS UI do inpainting?
No. It has start images, reference images for Flux.2 and Qwen-Image 2.1, and SeedVR2 upscaling. For inpainting, Krita AI Diffusion or InvokeAI are the better tools.
Which models does Krita AI Diffusion support?
Its docs list SDXL, Flux 2, Z-Image and SD 1.5, and Anima since 2026. Custom ComfyUI workflows can run others.
Is Krita AI Diffusion free?
Yes. The plugin is open source under GPL-3.0 and free to run locally. Its cloud service is paid.
Sources: Krita AI Diffusion on GitHub, releases, installation, ComfyUI setup, HEISS UI README.
