First published model, so it's guaranteed to have some issues. Kinda got curious and decided to try my hand at making a LoRA. Trained using 349 images from Onee-san Sensei, Kajuu 100%, and Onee-san X Shuffle.

Probably over-trained, but it looks like it works decently. Hands can get a bit wonky depending on the pose, and the LoRa has some... quirks. Might need to up tag strength if you want certain poses/hand gestures/hand positions.

This LoRa isn't designed for generating the characters, just meant to emulate the style from the training images.

Use with 0.8-1 strength.

Made using Civitai's trainer.