Selena - Knightly Passion

I still haven't found the optimal way to create a dataset. The dataset contains 88 different images and their captions. Some have collarbone piercings, some don't. I wanted to reproduce them, but it's probably not detailed enough.

She primarily listens to the following set of words: vamp1re, 1girl, silver hair, red eyes.

She also optionally listens to:

collarbone piercing, finger rings, thigh straps.

I didn't add the rose on the head to the trigger word, if you want to see it, just use: rose hair ornament.

I also don't know the optimal number of steps to ensure everything is detailed without overfitting. Therefore, I'll add several files with different step counts.