This is the first lora I have trained, its sort of over baked but it is consistent :) I would recommend 0.6 or 0.7 when using it.

It doesnt need any trigger words although all images were captioned with full tags and had the word "bankyu" you can just prompt with a general pos+neg prompt along with 1girl and the results come out looking like her. You can obviously add stuff like "white skirt, red glasses" if you want those features to be more prominent or "bankyu" if you want an overall strength increase.

Trained using AOM2 and some random settings I stole from a youtube video.

I have absolutely no idea what I am doing.