This LoRa was trained with 60 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:
https://www.pixiv.net/users/7069151
SuperSoftSon (@SuperFantoasts) / 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.
Description
FAQ
Comments (4)
Why 217.89 MB?
It’s the standard 32-dim size, no?
@Mexes Yes, but you made https://civitai.com/articles/23726 and I thought you couldn't make the mistake: Making the rank too high
@qek Often, you stick more with practice than theory. This works for every LoRA I make; I could risk lowering it to 16 or even 8, but in my opinion, 32 dim is the sweet spot. I mentioned that in the article.