Petite v0.3

I trained this model off of generated images of women who were under 150cm/5ft and 25-35 years old and a mix of SFW and NSFW themes. Most of them were significantly under that height, however in practice it doesn't remove more than a few cm/inches from the average height. I believe it needs more training to get right, but an artifact crept in on two different trainings with different sets of images. If anyone knows how to get rid of the swirly light/hair thing going on at higher strengths I'd appreciate it.

A little thought behind how and why I trained this LoRA:

I used full images, no blank backgrounds. Since a woman is the most common training an Model has, I've taken the perspective that it's the most baked and therefore the hardest to change. If I just use tall men next to short women, since it's less sure of what the average man looks like (fewer training images in the base model) it's more likely to change the size of the man to be extraordinarily tall. By having multiple things in the image that are standard sizes, like doors and refrigerators, as well as men I can push the perspective that the woman is shorter. That's my theory and the reasoning behind why my training data looks the way it does.