lora train on various images (anime/AI/realistic) to force it to be compatible with several models and graphic styles.

Perhaps you would prefer to use a basic template designed for cartoon style?

Have fun.