I asked grok to describe a reference Korean beauty, and then, based on this description, I generated a dataset, which I used to train LORA, using the main points of the description as trigger words. You can see the result on your screens.
Verion 2
I completely updated the dataset, I wanted to achieve greater compatibility with a variety of styles without losing realism, it turned out well.

Description
I completely updated the dataset, I wanted to achieve greater compatibility with a variety of styles without losing realism, it turned out well.



















