Lora trained in recreate Mochi (Circle Rin)'s art style.
Anima Preview v1:
I was able to train LoKRs for this model and they keeps the styles better than loras (as it was on Stable Diffusion Models). I'm not sure if I did a correct implementation of this network, but it works without any problem in Comfyui. Tell me if you have any problem.
Weight: 1
llustrious01 v2:
I re-trained the model due I feel it a bit under-backed, now is a bit over-backed xd, but it helps to keep the style when is used on derivated models.
Weight: .8 - 1
llustrious01 v1:
Weight: .8 - 1
Pony6 v1:
Weight: .8 - 1
Description
FAQ
Comments (5)
You are God.
Haha, not really, I leave a bit under-baked this (and other) model due I was using a new config for train them. I'll fix that as soon as possible. (I still want do some tests before train a model with a large dataset). (this is just with the Illustrious version)
I used the "Vpred" checkpoint to test it, and there was no problem with the style, and some of the movements had anatomical and perspective flaws. (The machine translation used may not be easy to understand, sorry)
@Black130516 Now you can see and compare what I said about the under-backed problem.
@Konoko You really are my god. XD