V2 is a much cleaner version trained on edited mostly text-less images.

Add "censor bar" to your negative prompt, since source images are often censored.