First attempt to train a LoRA on something other than 1.5 (trained with AnyLoRA). Tested with multiple anime and realistic models using just very short prompts (subject, trigger). Works with other character LoRAs without breaking their style but of course you need to play with weights on both to find the best compromise.