It is meant to make images look more like amateur studio photos. The model has been trained with a dataset that contains both sfw/nsfw using ai-toolkit.

For images side by side, left is without the LoRA and right with LoRA at strength 1.

I found Euler simple at 20 steps gives best results. 

Resolutions it was trained on.

704x992: 1 files
704x1248: 5 files
832x1216: 55 files
1152x768: 6 files
896x1216: 4 files
768x1152: 16 files
736x1088: 4 files
1216x832: 2 files
1120x1440: 23 files
832x1152: 2 files

A combination of short and long prompts were used.

This LoRA is not perfect. It has a hard time doing night shots and can sometimes create distorted limbs.