LORA made to generate images featuring female gymnasts. I trained this on 136 images. It's still incredibly wonky, especially since SD has trouble with poses more complex than standing straight, but there is definitely room for improvement.

Keywords:

• Gymnastics outfit: leotard

• Gymnastics ribbon: gymnastics ribbon

• Standing splits/I-balance: standing split

• Gymnastics ball: ball

I also planned on adding more keywords, but they barely work. Looks like I'll have to train this model for a little longer ¯\_(ツ)_/¯

All example images were generated using MeinaMixV10 checkpoint (this LORA isn't doing it justice tbh) and EasyNegative textural inversion.
https://civarchive.com/models/7240?modelVersionId=80511
https://civarchive.com/models/7808?modelVersionId=9208