They're using a previous generation of TTS models, which most of the reader apps are using. They're reasonable, cheap, but sound noticeably worse than OpenAI's or 11Labs. I don't like them.