This seems like something that should be handled on the computer side of things. Learning to use one dictation system and make it work seems hard enough. Having a different one on every piece of software you use seems like it would be terrible. Trying to support dictation on multiple platforms through the browser also seems untenable.
Wouldn’t you rather see that the voice-to-text actually worked before it was uploaded to a forum? Doesn’t voice-to-text work better on your own computer where it can be trained to your own voice?
I’m all for supporting all kinds of users, but I don’t see how it makes sense for every single web application to be responsible for transcription. We don’t expect Discourse to read text to the blind, Discourse provides HTML that the application that the user chooses can work with.