Sharing my lora with you, it works with many of the models available, even photorealistic(horse_penis_V0), v1 wasn't tested on photo realistic models.
I was doing some tests with v12,v13,v13a with some 3d renders related to rear view, which works fine in v13a, the other versions work sometimes, I'm planning to add first person view.
Prompts without negative were made for testing purpose only.
Rear view should work fine with prompts like "futanari, 1girl, ass focus, ass, cum in ass, anus, testicles, horse penis " with "from behind, rear view" at the end.
Because it is trained using clip 2, on models trained with clip 1 it tends to produce strange results, however it is possible to make it work using other layers,
Negative embed are from the links below
https://huggingface.co/datasets/gsdf/EasyNegative
fixing some things, focused mainly on the from below and close-up tags, changed some parameters that allow longer prompts which allows them to work better, changing the dim from 36 to 64, which resulted in a bigger file
Go ahead and upload yours!
Your query returned no results – please try removing some filters or trying a different term.