I've found that Google's speech to text works surprisingly well for me, and I'm not the very best at clarity and enunciation.
I generally know what it's going to mess up in advance so I can just fix that.
Of course, this isn't very useful for an aurally communal work environment like most offices.
Can you just send speech directly in this situation, without converting to text? There are apps, that allow asynchronous speech messages.
It's aggravating. But then later when I can continue a conversation via keyboard the depth and intimacy is much greater.
Maybe as speech to text improves it'll get better.