Brief
This is the lora model for Raiden Mei from Honkai impact 3, post-honkai odyssey, i.e. the adult version of our lovely Mei！Trained with 1024x1024 resolution, and enhanced with 3D model for best accuracy.The training dataset was categorized to further improve the flexibility, making my model easy to remove/change the costume.
Use:
V1, V1.5 and V2
Trigger prompt: raiden mei
Default costume: pantyhose, boots, single glove, earrings, dress, choker, tank top
Weight: 0.7~1
V3 and SDXL
Trigger prompt: raiden mei
Default costume: pantyhose, high heel boots, single glove, earrings, layered dress, ribbon choker, tank top, detached sleeves
About training
V1 was trained basing on anything V3.0, V1.5 and after was on NAI.
SDXL version was trained on animage 3 base
Known issues
V1:
Mei's costume is too complicated, especially the feathers and the purple pattern near the chest area, so it might get wrong. The necktie is also too difficult to remove. I will try to fix them in the future versions.
V1.5:
Chest area and collar feather for the default costume
V2.0:
Sleeves and the single glove might not be correctly generated every time
Description
Improved face
Slightly improved accuracy of her default costume
FAQ
Comments (3)
finally, he change the clip skip to 2.... i mean your Loras are really good... but please don't make it clip skip 1, i waited so long for this Lora change to clip skip 2
But you can always set the clip skip as you want, right? Is there a API thing? Maybe I can recheck and set my other models to 2 too. BTW, I set it to 1 is because the training scripts, but when comes to generation, 1 or 2 does not make too much difference
in that case i can only use this Lora alone, because most Loras were trained with clip skip 2, so if i mix this lora with another style lora or detail lora it became unstable