This was a first attempt and the results are not always perfect. It turns out to be a very difficult topic to train on and the idea was just make wan slightly better than what it is.
Depending on time and motivation, I might improve it in the future.
Captions were done with JoyCaption for images and by hand for videos, so that should give you a hint on how to prompt.
Two sample prompts:
The scene is a close-up featuring a woman's vulva, the subject is filmed from below. The subject is a woman with red hair, she is nude and has a ring on her finger. She is moving her fingers slowly across her wet labia. The scene shows pussy play. This close-up photograph features a woman's vulva, with her labia and clitoral hood clearly visible. The skin tone is light, with a natural, unfiltered appearance and visible pores. The lighting is soft and natural, likely from ambient indoor light, providing a gentle illumination that highlights the skin's texture without harsh shadows. The camera angle is slightly tilted, focusing directly on the genital area, with a shallow depth of field that keeps the background blurred. The background appears to be a beige carpet, adding a neutral, non-distracting element to the composition. There is no visible makeup, and the skin looks natural and unaltered. The aesthetic quality of the image is moderate, with an emphasis on clarity and detail rather than artistic composition. No trigger word.
Description
The lora was trained for 3000 steps in images from a place you might know. The prompts are just filler because of civitiai issues
Details
Downloads
5,380
Platform
CivitAI
Platform Status
Available
Created
9/14/2025
Updated
12/16/2025
Deleted
-
Files
ps_high_noise.safetensors
Mirrors
Huggingface (2 mirrors)
CivitAI (1 mirrors)