Take 2: This version used an entirely different dataset, with the same goal. You shouldn't expect many practical differences. Fewer random navel piercings is one benefit.

Same blurry line problem from dataset resolution, easy fix, for those who are interested.

The goal of this LoRA was to recreate the style of the casual but intimate adult modeling done by several photographers and networks online.

Strengths

• Natural faces - not ideal plastic supermodels that all look the same

• Authentic settings - lived-in interior indoor locations and banal outdoor sets

• Details - vellus hair, freckles, pores, piercings, goosebumps all seem to come out natural looking

Drawbacks/Problems/Stupid stuff

• There is almost always a thin blurry line running along the bottom of the image. I think this might be due to how I cropped images when making the dataset. It can easily be worked around. If you can share advice on how to fix this for the next generation, feel free to let me know.

• Trained with few concepts, so you will need to use other LoRAs to get hardcore things. All of my sample images used FilterBypass, as well as SNOFS at low weight.

Why early access?

I used to have buzz. Now I don't. I made this LoRA so I can download other EA models and test them myself.