Sardine184 trained an Anima version of this LoRA using the same dataset (or at least I //thiiiiink// is the right dataset). It might struggle with wet and messy diapers, but try playing around with the LoRA weight and the prompt(s). https://huggingface.co/sardine184/Pullup_diaper_Anima_V1

Illustrious LoRA for characters wearing "pants-type diapers" (or "pullups")

This model is intended for anime/manga-style image generation with "pants-type diapers" (and often called "pullups" by us westerners). That being said, this version was trained on a greater variety of art styles, including more realistic ones. Unlike earlier versions, only white pullups were included in the training set. Print pullups might still work, but not guaranteed.

This LoRA was trained on Illustrious 0.1, but the images posted alongside this model were generated with Nova Anime XL. That said, I tested a "beta" version of sorts with checkpoints that use more realistic lighting / shading, such as Nova Unreal, with pretty good results. For best results, a low CFG scale (like 2) is advised.

No trigger word is required for this LoRA. Add "diaper" for a dry pullup, and add "wet diaper" to the prompt for a wet diaper. For a messy pullup, including "soiled diaper" in the prompt may yield results depending on how high a guidance value you give to the LoRA.

If you already have ComfyUI installed, and have Nova Anime XL v8.0 downloaded and added to your checkpoints folder, I have posted an example workflow you can play with (that currently generates a pic of Shizuka from Takopi's Original Sin... kiiiinda). You can find it among the files as "Example_ComfyUI_Workflow.json" (If you don't have Nova Anime XL v8.0 downloaded yet, you can get it from https://civitai.com/models/376130 - selecting "IL v8.0" on the page. You could also choose to download the newest version and just modify which checkpoint the workflow uses.)