Little things about the LORA itself unlike previous I highly recommend using the tag short hair since for some reason it likes trying to give her long even though none of the training data had long hair, and it tends to bounce to brown or black hair for her (understandable problem honestly) put the tag for the preferred color if it bugs you. Okay so this had a lot of changes involved and there was a lot more in between than simply V1 to V2 I learned a good bit, got completely lost and then came to a new compromise when developing this LORA. As I recently learned you do not need to crop training images (learn something new everyday) up to now I was using 512x512 cropped images so this fact was revolutionary. Also why playing with the formula I learned how to make the LORA a small data size taking it from 148 kb to 19 kb. Also I tested training in NAI again and this time all my pictures weren't super dim (probably because it had bigger images) so this should be compatible with other anime LORA better than before (Once I remake some of my clothing LORA I will test it more but it looked good when tested by style LORA). If anyone has any ideas or suggestions to continue improving or just curious how I did anything feel free to make a comment. I'll probably just be updating my current Lora to this new format and then I'll end up remaking the arcade LORA at the end, again (the one that started this update mission to begin with.
*Credits to 10komi for the senran and lykon for the 90's
