State of the art text-to-speech for 2025?

I'm aware of OpenAI's text-to-speech model, and ElevenLabs' text-to-speech models are very good as well -- what else is out there? What else is approaching that level of quality and diction?