I'm using the Speakable Items feature in Mac OS X, to voice activate some simple stuff like "Play Music", etc.

Since I'm in a lab and put my macbook off to a corner of my desk, connecting it to a monitor and other I/O devices for ergonomic reasons, I had to yell to get my voice to reach the built-in microphone. My labmates were not happy, understandably.

To fix this issue, I switched to a bluetooth microphone that I clip to my collar. I no longer need to yell to reach my macbook's built-in microphone. However, this also means that I have to retrain the speech recognition engine as the audio signal it now gets is very different.

Here's the problem: When I open the calibrate window in the Speech prefpane, it performs some sort of unsupervised learning. I can't even tell it what I'm saying. This means that I have to guess what it thinks each command sounds like and then repeatedly say that command, changing the way I sound each time; until I sound the way I normally do (accents, volume, pitch, etc). This is super frustrating.

Is there a better way to train the speech recognition engine? I don't seem to be able to train it quickly enough (been at it literally all day) and I have a ton of custom speech commands that I am unable to use because of this fail.

Technical specs:

  • Mac OS X 10.7.5 (Lion)
  • a good bluetooth microphone (Sony)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.