This LoRa was trained with 88 images, which were cleaned of text.

I try to make sure the results you see are the ones you’ll get. The examples don’t go through any processing (hires fix, adetailer, inpaint, etc.), so it’s likely you’ll notice some errors in them.

Please, if you think the style doesn’t match or if there are mistakes, let me know — I’ll be more than happy to review it.

Do you have a style request?
Leave it in the comments and I’ll make sure to write it down.

If you like the style, support the artist on their official social media:

Just Umbra - pixiv

Umbra (@krokobyaka) / X

If the artist requests it, this LoRa will be removed.

Made it this far?

If you like the model, let me know! Comments, feedback, suggestions, even insults—everything’s welcome.

Also, feel free to explore my model library — I’ve got many LoRAs in various styles that you might enjoy.

And if you’d like to support me, you can do so on Ko-fi. Even the smallest contribution means a lot and truly motivates me to keep making models and tools for the community.

krokobyaka / Just Umbra | Artist styles | Mexes - v1.0

krokobyaka / Just Umbra | Artist styles | Mexes - v1.0

Sampler

Steps

CFG Scale

Seed

Clip Skip

Size