The platform is actually missing two important parameters: the speaking style (I suppose it’s always defaulting to General - BTW, on the API this seems to be the “type” parameter) and the Language Skill.
For instance, there are some voices that are able to whisper (good for voice notifications triggered very late ), but only if those parameters can be set.
There is a pending pull request that bumps up the version of the underlying library for the Microsoft TTS service. If that PR was approved, you could embed Microsoft’s SSML which would let you add in special XML that configures the speaking style and many other settings. But that PR has been stalled out for a long time.