Here I leave a small introduction of the LoRA:
I like how it works together with Power_Mix.
I'm working on a more complete and much bigger version, it should be available next week, this version is just a test for IL v2.0 7w7)/
This version attempts to maintain the coloring while enhancing the faces and adding some detail. The general idea was to use art that was quite distant from the original dataset due to the differences in the base models (i.e., I couldn't find the original dataset and was too lazy to redo it).
PS: This is a style based on my generations (I couldn't find the original dataset and I think it's unnecessary).
I was reviewing the dataset of this lora because after copying a prompt something appeared that should not appear... and I realized that I used the wrong folder, they are practically the same files but to train styles with pony it is necessary to be quite specific I mean, if many images have a name that is repeated, it can cause the lora to classify it as a set.
In this case "kimkrab" was not completely deleted from the dataset so you can use it as a trigger but this lora has more than three artists and I doubt that I forgot to delete any of their names. (or add it, after all there are only names when I plan to do loras of a single artist).
PS: Tatsumaki is hungry, give him love ~
Well, I have no idea why it wasn't working in the site generator, the solution was to delete some metadata (doesn't affect the Lora)... P.S. This version includes improved hands and smooths out the style. It was also trained in "cosine with resets" unlike the previous "constant" version. There are some new insights :3
1. Quyền đối với các mô hình được đăng lại thuộc về người sáng tạo ban đầu.
2. Người sáng tạo gốc muốn xác nhận mô hình vui lòng liên hệ nhân viên SeaArt AI qua kênh chính thức. Nhấp để xác nhận
