LoRA trained from screenshots taken from the currently available episodes of Season 4 so far, for anyone who wants something that stays pretty close to the anime’s style.

There are still some details I’m sure could be improved, but I’m too lazy to keep retraining it over and over again. Maybe I’ll revisit it once the season is over and there are more images available.

The base model used for training was Illustrious-XL v0.1. The sample images were generated using animemix v8.0, though I’m not sure how well it works with other models.

Prompt for normal version:

emilia-s3, purple_eyes, grey_hair, long_hair, crown_braid, large_breasts, pointy_ears, elf, hair_ornament, hair_flower, white_flower, hair_ribbon, purple_ribbon, x_hair_ornament, white_cloak, hooded_cloak, animal_hood, hood_down, purple_bodysuit, (white_boots, purple_trim)

Prompt for the version with the hood up:

emilia-s3, purple_eyes, grey_hair, long_hair, large_breasts, x_hair_ornament, white_cloak, hooded_cloak, (animal_hood, purple_ears), hood_up, purple_bodysuit, (white_boots, purple_trim)

Prompt for version without cloak:

emilia-s3, purple_eyes, grey_hair, very_long_hair, crown_braid, low-braided_long_hair, large_breasts, pointy_ears, elf, hair_ornament, hair_flower, white_flower, hair_ribbon, purple_ribbon, x_hair_ornament, crop_top, detached_sleeves, purple_bodysuit, (white_boots, purple_trim)

I don’t know, this is just how I use it.
I’m not really sure what people might use the LoRA for either, I just felt like uploading it and I did it.