All modern models have their default looks. Meaningful variety of outputs for the same inputs in finetuned models is still an open technical problem. It's not impossible, but not solved either.