This is a very very early pre-pre-pre-release of the model. I am already training a new version, which will be published soon, but I like this one. Even though I made quite a few mistakes preparing the dataset, most of the mistakes I made, produce quite interresting results.

Recommended settings are:

steps: 40...60,

cfg: 6...13,

sampler: dpmpp_3m_sde_gpu,

scheduler: karras

The model can be used with or without the SDXL refiner.

Since this is my first public release of a LoRA, I would appreciate any feedback that will help improve the model. I am also planning to train a checkpoint soon. I am a datascientist so this is a part of my work and I have fun doing that :)

You can feel free to do whatever you want with this model and the licence terms are as wide as possible on this platform.

All the images I used for training were tagged with keywords, so you can just plug the model in and it will work. But is you want more you can use these keywords:

beautiful face, beautiful pussy, beautiful tits, naked woman, transparent cloths, wet body, wet cloths

Please, note, that the model is NSFW.