DIY: Speech-to-text alternatives to Dragon NaturallySpeaking

If Dragon NaturallySpeaking speech recognition software isn't in your budget, read about other speech-to-text options Jack Wallen suggests to a TechRepublic reader.

I received the following question from TechRepublic member Jeff Braddock. Please post your suggestions for the reader in the discussion.

Q: At our church, we have a deaf man, and I would like to make it where he could participate in/enjoy the services. In particular, I am searching for some kind of real-time speech to text program. I know there is Dragon, but 1) it's not free, 2) I don't know how well it would do.

I would be open to something that either runs on Windows XP or a version of Linux. I currently dual boot Windows 7 and Ubuntu 10.04 on my laptop, and so would be able to test it out first.

I have tried using the speech to text capabilities in Word 2003, but even with several training sessions, it just isn't accurate enough.

I/we would appreciate any suggestions you might have (even if they are not free, though that is always nice! :-)

A: As you may know, Windows (both XP and 7) has built-in speech-to-text recognition. It has some quirks and faults, but it can do the trick. You might be running into incompatibilities between versions. Here's what I know:
  • Windows XP does not support speech recognition in Microsoft Office 2007 programs.
  • Microsoft Office 2003 and 2002 works fine Windows XP.
  • Microsoft Office versions prior to 2002 do not support speech recognition.
  • With any speech-to-text program, it's important to train the tool for the voice it will be hearing, so make sure the speaker gets plenty of training time in a quiet space with a good mic.

With that in mind, I recommend using either Windows XP and Office 2003 or 2007 or use Windows 7 and Office 2010 or 2007. If you use any other combination of OS and Office suite, you're going to have issues.

In Linux, there are two options: CMU Sphinx and Julius; however, both options are for users with quite a bit of experience with Linux and can be problematic to piece together. If you're interested, here are more details:

  • CMU Sphinx is a collection of tools for people who work with speech recognition. It's written entirely in Java, so the installation might be a challenge.
  • Julius is an open source large vocabulary CSR engine. Make sure you spend some time on the website — you might find that Julius is just what you need.

Keep in mind there's a reason why Dragon NaturallySpeaking is so popular, as well as pricey — it works.

Ask Jack: If you have a DIY question, email it to me, and I'll do my best to answer it. (Read guidelines about submitting DIY questions.)

About Jack Wallen

Jack Wallen is an award-winning writer for TechRepublic and He’s an avid promoter of open source and the voice of The Android Expert. For more news about Jack Wallen, visit his website

Editor's Picks

Free Newsletters, In your Inbox