This does not currently answer the question, but I have been working on a personal project since 2011 for this very purpose. While it isn't finished yet, I estimate it to be completed (and by completed, I mean ready to be used - but probably needs a lot of testing with many different accents) within the next month (give or take a little).
In its current state, I am able to speak-out an entire app (a small app - for testing purposes). Let's take this app for example:
static void Main(string args)
The way you would 'write' this app using voice, would be to say it like this:
Namespace ConsoleApplication 2. Class Program. Static void Main takes
1 parameter called args, of type string array. Inside Main, say Hello.
I have no plans to support any other languages in the near or distant future, but when the source is released, I would love to see other languages being supported. I am trying extremely hard to prevent the need to say any symbol names, because they totally break the normal flow of speech and make it much more time-consuming and difficult to voice-code. This is the main reason for only working on one language; it's just too time-consuming. And I need to be able to get support for one language right, before I even bother with another.
I didn't plan on 'releasing' the program, since it was made to fill a need of my own, but I just came across your question so I might as well release it. I have setup a GitHub repo for it, and the very second I can see that it is reliable-enough to be released, I will push to my repo and upload the binaries.
The repo can be found at: https://github.com/SE505/SpeakSharply
I'll also keep an eye out to see for any changes in this question and I will update this answer. I'd love to see some new answers (and software).