Trouble is, it’s still the same model, just with a request to sound different. It can only go so far. It’s much like asking an author to imitate the style of another author. They might do a decent job of it but they’ll never match perfectly. And that’s for humans who can actually learn. LLMs can’t.