LoRA is generated entirely from anime images.
Should work well with any model.
She's my no.1 waifu, so I tried to give her more time when I was preparing her dataset.
Check samples for prompts :)
Leave a like, comment, review ( with new images ;) ) but first of all, ENJOY ;)
If you wish to support me in my work,
you can do that here: Ko-Fi.com/LittleJelly
Description
FAQ
Comments (11)
example images look great. will try ^_^
Nice work! Did you use any Hires fix in your samples?
Always.
HiResFix from: 512x768 to 1024x1536 + Upscaler 4x-AnimeSharp (upscale by 2)
I would like to say thank you from bottom of my heart!!!!
this looks amazing, can you please do luvia some time as well?
Uhm, where did she appear a lot? Because in UBW there were only like 2-3 scenes with her and thats not enough to generate a decent LoRA :/
@LittleJelly There is a lot of her in the Lord El-Melloi Case Files
@JONDOOM1811 I have Lord El-Melloi on my ToDo list, so I'll give it a go
@LittleJelly cool, another option is the version from Kaleid for Luvia, and even Rin, I honestly find the design pretty cute of the anime
whenever I try to use high res on this model, I get "NansException: A tensor with all NaNs was produced in VAE. This could be because there's not enough precision to represent the picture. Try adding --no-half-vae commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check."
what do I do?
It's not a LoRA problem. That's a VAE problem.
Solution: Use a different VAE.