From text to speech – why the future of AI is about the voice