The problem is, even if you do build amazing speech to text, it will br slower and less expressive than a keyboard + pointing device (mouse, touch, pen).

For keyboards, you lose positional logic (wasd in games). You lose shortcuts. You lose control over capitalization and formatring. You lose punctuation. You lose non-text input (code, dictating code sounds like like a horrible pain). You lose function keys. And, of course, you lose speed (think of instant things you do with shortcut keys, like alt tab). Not to mention, that you lose the ability to work in silence.

Make the recognition quality gorgeous, and it will still be a less flexible product than what we use today. It has value for accessibility, but people will likely choose keyboards over dictation based on UX alone.