You can reduce the time that recognition happens by changing the commands that you are using to, 'Restricted Continuous Speech' for the, 'Recognition' option (on the Add/Edit a Command screen). Something else to look at is making sure your, 'Recognized Speech Delay' on the Options > Recognition tab is set as low as possible (the default value is zero, I think - leave it at zero if you havent changed it
).
Also, a properly-trained speech engine helps a bit, as well as a quiet environment.
Hope that helps!