As an experienced Dragon Naturally Speaking (DNS) user - I'm using it now to dictate my reply to this thread - I would have to agree that this is not the solution for you unless you want gibberish. DNS works well if you optimize it, but it takes quite a bit of time to get the highest accuracy out of it.
When I needed some audio converted to text, around eight hours worth, I put the job up on rent a coder (now freelancer), and I was amazed how little it costs me to get it transcribed with a 100% accuracy. It is my recollection I paid around $30.