We’ve trained a Large Language Model that is capable of generating speech with true human-like expressiveness and realism.