I use it,
The biggest advantage is that it takes me from 15 WPM (hunt and peck method) to 125+ WPM. Also the software is selective enough to "Not" type when my wife is speaking, allowing comments from her as I "write".
The biggest disadvantage is that The software is not self aware, and will put in what the algorithm "Thinks" you said. Our most famous is where I was writing an article on Steam Engine design, and I said "Springs and levers" The software wrote "Friends and lovers" (secret railroader in-speak?)
The software allows me to get the story down "on paper" and edits are absolutely needed. For me the hardest part is getting the first draft completed, and the software helps tremendously. Also of note is that manual inputs are still recorded while you are speaking, so line breaks and punctuation can be inserted as you record.
I know that both David Weber and John Ringo use voice recognition software, and that it allows them to greatly increase their production.
Regards,
Kevin
Naturally Speaking is the one used at my university.