V1 - (Failed)

This model aims to help in the generation of images of Fumino Furahashi from Bokutachi wa benkyō ga dekinai.

Sadly, i have to admit that the material i trained on wasn't that good, and the result is mediocre, but i think it might have some use.

Let's talk about the model, it was trained on highres detailed images and this made the lora have an impact similar to "add details" lora.

The activation tag is:

FuminoFuruhashi

It is optional, but strongly recomended, the use of:

white headband
long hair        <-- when the hair are tied up, this fix the issue.
mole             <-- or mole under eye

(The model sadly didn't recognized the right position of the mole, but it fix instantly with inpaint)

The model works with most outfits (and it has not a standard outfit), but has some bias you should know:

It has a bias towards blue bikini, when the tag bikini is used
It has not enough data on green outfits, and often mess up the colors of those
It support nsfw, so it might turn a regular sweater in a "virgin killer sweater"
It support nsfw, so you might wanna use "nude" as negative prompt, to avoid unwanted nsfw art

The suggested settings are:

Weight: Between 0.6 - 1  (i often choosed 0.66 or 0.75)

It works good with euler a / dpm++ 2m sde kerras

Tags that could improve the generation are:

Bokutachi wa benkyō ga dekinai
Fumino Furuashi

It often overdetail the character so you might wanna use add details with low (or even negative) weight