Ask HN: What is the state of the art in text to speech? Has much changed?

Last time I closely examined text to speech was 2-3 years ago. At the time, Google, Microsoft, Amazon and IBM had okay text to speech but nothing that was super impressive. Certainly you could always tell it was read by a computer. The best at the time were the "neural voices" offered by Google and IBM. I'm wondering what has changed since then? What is the current state of the art in text to speech? I would have thought that AI might have completely changed the landscape? Can anyone suggest any really impressive modern text to speech systems?