Tags: patikitten, anime girl, green eyes, cat ears, closed mouth, white top, collar

T_T I tested blip captioning to see how good the accuracy would be compared to danbooru tags

It has issues with the jacket, hair color, bandage & fish neck collar due to not being tagged properly

The LoRA is not flexible due to tagging errors 😿

It actually seems to be more accurate in autismmix than pony weirdly even though the base model is pony.

Training data (taken directly from vtube studio): https://imgur.com/a/bfgSzna