Style lora for creating anime screencap-like images.
Use "screencap" in your prompt. This word was not actually used in training but it seems to work well.
The training data was biased towards upper body/face close-ups. Prompt for lower body related things (pants, skirts, legs, etc.) to get more of the body visible.
(The details in metadata are a bit wrong. Ignore them. Forgot to update them after copy-pasting training settings from somewhere else)
Training details:
The model was trained from 571 manually selected and cropped screencaps from yozakura quartet.
The images were captioned with wd-vit-large-tagger-v3. Minor manual fixes were applied to the tags.
V1.1: Trained from 768 images and more epochs.
I noticed that I had actually prepared 768 images for the training. Somehow I had accidentally not used them all.
V1 :The model was trained from 571 images.