Couldn't find an issue for this so I'll be making one:

Currently seemingly the best voice model is from Google, Google Gemini 2.5 Pro & Flash TTS - And it's good in one specific way - it supports more niche languages much, much better, specifically Lithuanian. It would be awesome if support for it could be implemented.

On a side note, Google also has a realtime API for speech but it's probably much harder to integrate into a package like this but would also be pretty neat to have in this package. The realtime APIs are not for text to speech but for natural voice modes like on the chatgpt or Gemini apps.