Current Environment:

Using speech recognition software: tips and tricks for dictating

How does it work?

It is important to understand what is going on when you use speech recognition.  The simple explanation is:

  1. You push air out of your lungs
  2. Through your vocal cords
  3. Out of your mouth comes sound waves
  4. Those sound waves hit the microphone
  5. The microphone converts the sound waves to an electrical signal
  6. The electrical signal travels from the microphone to the computer
  7. The software looks at that electrical signal and compares it to the thousands of electrical signals in its data bank
  8. The software makes its best match
  9. The software puts the best match up on the computer screen

Remember — the software is simply a matching program. There is no brain, no understanding of what the words mean, and no understanding of what you are trying to type. The software is simply matching electrical signals.

So if you want the software to type what you say, you need to say the words in a way that will match how the programmers put the words into the data bank. Do not over enunciate. Do not yell. Do not put extra emphasis on the last consonant. While these things may make it easier for the human ear to understand your speech, it makes it more difficult for the computer.

Clear, relaxed, ‘crisp’ speech is best. 

If you have difficulty with breath support, try speaking only three or four words at a time instead of a whole sentence. If you run out of breath while you are speaking, the words at the end of the sentence will sound very different from the words at the beginning of the sentence. Consistency is important for the computer to learn your speech.

That means, instead of:

“Now is the time for all good men to come to the aid of their country period”


“Now is…the time for…all good men…to come to…the aid of…their country period”

If you speak with an accent — when you set up your voice file, you will be given an opportunity to pick the appropriate accent. This will make a big difference in how well the software will be able to type what you say.  Also, be extra conscientious about correcting software recognition errors to give the program a chance to adapt to your voice.