⚠️Important: Each LoRA version is designed for a different base model. Please make sure to download the version compatible with your base model.
This is a LoRA for xipa(@jipa42760710)'s recent Style, created on request.
I will remove this if requested by the original artist.
Trigger words:
For v2.0 AB1: @xipa2026late03-05,
For v0.1-1.1 Anima, A2.9B, Moss: @x1p4early2026,
For v0.01(Krea2 test): xipa style.
Note:
For v0.01 Moss: Due to an insufficient number of training steps, you will not get the expected effect unless you significantly increase the LoRA weight. For your reference, a weight of 1.5 was appropriate in my image generation environment.
Description
His art style in January 2026 to mid-March.
Trained on Anima Preview v02.
Reduced the occurrence of unintended skin color variations.
FAQ
Comments (4)
It would be amazing if you made a version of Cyber-Wifu11
Hot af~
这个lora实在是太赞了。
想问一下楼主用的训练参数是什么，我现在在训练角色但是最后效果很差，用了cosine和constant的schedule和warmup，1e-4的学习率，但是出来的lora都没有光辉训练的lycoris要好，anima在训练的时候是有什么需要额外注意的东西吗
Thanks! And regarding the character LoRA, which is the main topic, I haven't been able to make them well either. I'd like to know the good settings too.