Show HN: Krita Stable Diffusion Plugin

[_7eg9]

This is still a very new project under active development but I wanted to show it off in case anyone is interested in bookmarking it or contributing.

Demo here: https://www.youtube.com/watch?v=maWR7dDf4SE

Although it will be improving, the codebase has a few issues

The Good

  - ability to multitask within krita while generating images

  - uses a queue so you can queue up multiple images without waiting for your other batches to come back

  - no need for a webserver (uses sockets to communicate)

  - clean code

  - active development (i want to use this tool myself)

  - uses stablediffusiond and stablediffusion as separate repos, these can be swapped out as desired with some hacking

The Bad

  - currently relies on rabbitmq and doesn't provide another option

  - plugin only does txt2txt. img2img works but is disabled

  - must acquire the model manually

The ugly

  - not all of the features work yet. in fact you can only generate an image.

  - installation is somewhat involved (requires CUDA 11.3, rabbitmq and more) - there is an installation file and instructions but likely need improvement,

Krita Stable Diffusion plugin

  https://github.com/w4ffl35/krita_stable_diffusion

stablediffusiond (daemon / queue runners)

  https://github.com/w4ffl35/stablediffusiond/tree/feature/krita-plugin

stable diffusion

  https://github.com/w4ffl35/stable-diffusion/tree/feature/krita-plugin

---

if anyone uses this and runs into problems, just open an issue in the appropriate repo and i'll do my best to help.

[fragmede]

Neat! Does this have any relation to https://github.com/sddebz/stable-diffusion-krita-plugin?

[_7eg9]

The only relation is that I used that plug-in prior to writing mine. I was inspired by it to make my own. I originally created a NSFW filter for that plug-in you mentioned. The code still exists but I closed the branch. No offense intended to the authors but...

The code in that repo is not in great shape. It is forked from stable diffusion webui and has diverged from the parent fork (at time of this writing "This branch is 473 commits ahead, 271 commits behind sd-webui:main").

I've seen multiple complaints regarding the code. It was hard to add the NSFW filter, and even as such the maintainer never merged by pr. Streamers can't use the plug-in without a ban risk on twitch.

My code is very clean and well documented. Adding new panels etc to the krita plug-in itself is extremely easy thanks to the classes i wrote.

Furthermore the existing solution uses a webserver, whereas mine uses sockets. Existing has blocking requests (click button, wait for image for 8 seconds or more) whereas mine is non-blocking (uses threads and a queue system so you can keep working in krita while it generates).

I realize mine is still in a broken state, but it won't be much longer. If anyone wants to contribute that would also be helpful.

[_7eg9]

Not really sure why this is down voted. Did I say something wrong?

[_7eg9]

I also added a discussion board in case anyone wants to try it out and needs help getting set up

https://github.com/w4ffl35/krita_stable_diffusion/discussion...