NEW VERSION

Click the link below to download an updated version with better front/back support, and dry diapers by default:

https://civarchive.com/models/1273506

Intro

This model is intended for anime/manga-style image generation with "pants-type diapers" (and often called "pullups" by us westerners). A fairly large image set was used, made up entirely of AI-generated images with diapers drawn on after the fact. The image set included a variety of poses, and included situations with characters wearing dry, wet, and soiled "pullups". A fairly comprehensive tagging job was done on all the images, but style bleed may still happen.

Usage Information

Include "pullup" and "diaper" in your prompt to trigger the LoRA. With a higher guidance value (such as 7), it should result in a character wearing a dry diaper. For a wet pullup, add "wet diaper" to the prompt (although a low guidance value will likely make the diaper wet anyway). For a messy pullup, including "messy diaper" in the prompt and using a higher guidance value should yield results, it's not guaranteed.

This LoRA was trained on Illustrious 0.1, but I would advise using Nova Anime XL for best results. Lastly, the tags in the data set included some easter eggs worth finding.

Addendum

Don't be too surprised if I end up training a "V2" in the somewhat near future (with a more balanced dataset, instead of just throwing the kitchen sink like I did here).