But you don't really have to retrain the whole model to modify it's writing style, fine tuning would probably be enough