Hyper Donut Anuses

Here's a LoRA specialized at depiction of the Derriere Bagels.

Trained on the NAI leak, should be compatible with many models with the same base. Used unfiltered e621 data while training since I accidentally lost my processed dataset.

The LoRA is biased towards shots, but what would you expect? Putting , shot, rear view in negatives helps diversifying poses.