The dataset is made of anime screencaps.

I didn't train specific outfits, as there aren't really any. The tag "tbase" might help with accuracy of the character. Otherwise, stick to the tags I have listed on CivitAI.

I just used the parameters i was using for PonyXL LoRAs. Please tell me if there's anything wrong (errors, artifacts, etc.), so i can try avoiding those in the future.