This model is trained on 200 original images, uses wd-tagger for image tagging, all tagged {TXT_PROMPT}, contained in uploaded training data.

This LORA model works well on various models, and still keeps some generality. For better performance, try something like:

masterpiece, best quality, 1girl, shibuya rin, long hair, solo, black hair, green eyes, necktie, skirt, school uniform, bangs, cardigan, looking at viewer, necklace, jewelry, grey skirt, <lora:shibuya_rin_v2_ssdt_lora32_epoch25:1>