Dragon Dictate 3 for Mac is packed with more valuable and productive new features than ever before including the Audio Transcription feature. Dragon Dictate 3 can transcribe your audio notes and magically coverts that into text for you. As part of our a week of articles on Dragon Dictate 3 lets see how the Transcription (previously known as MacSpeech Scribe) performs and we ask just how accurate is it?
Back in the “olden days”, if you wanted to take your spoken word that was in an audio file and have it changed into text you needed an application called Mac speech scribe. This was a separate application to Mac speech, the name which Dragon Dictate for Mac used to be known by. Thankfully these two packages have been merged together to provide audio transcription.
Training For Transcription
Dragon Dictate 3 Uses whatever profile you have trained to perform it’s audio transcription. so if you’ve done the majority of your training in a quiet environment, ideal for speech recognition, then try to emulate this as best you can when recording your voice. As with DragonDictate’s other speech recognition functions, it’s ability to produce an accurate transcripts depends on how well you trained your current profile. So as always it’s always worthwhile to take the time out to train the application. Each training segment takes no more than about 5 – 6 minutes and each one does result in improved recognition.
Calibration is a simple question of recording 32 seconds of audio using your iPhone or any other recording device. As long as the audio is clear and without too much background noise you should be okay. Once your recorded video is loaded you then get the chance to make sure that everything marries up between what is spoken and what’s been transcribed.
When recording your audio for the training process you might want to consider reading from the source rather than trying to make something up as you go along. As you can see from the Jewish in the screenshot above, until you are familiar doing this, recording your thoughts does take a little bit of practice.
Hovering the mouse cursor over the text DragonDictate 3 shows you what it thinks is either a sentence, paragraph or partial piece of text and lets you start the calibration process. if everything is all okay you click on the accept button on the text then turns green signifying you have accepted that the audio text has been transcribed properly.
If you aren’t sure of what you’ve said, Pressing the play button lets you listen back to that particular segment of the audio file. So if you aren’t sure what you said or you need to train or edit the text this is your opportunity to do so. (see my bootnote below)
And thats it!
This isn’t so much training but more DragonDictate three getting used to the levels and the quality of recording from your source. All of my testing was done with nothing more than my iPhone and the built-in voice recorder application
Once calibration has been completed you can then load up any audio voice recordings created using your voice and the transcription process gets underway.
There’s only one slight major flaw.
During transcription training it does say you can perform calibration test at any time yet, for the life of me, I’ve been unable to find out how to do this and believe me I spent the best part of an hour looking through every possible option and they just appears to be no way to redo the calibration without setting up a complete new audio source and going through the entire Dragon dictate 3 training process.
It works! But there are a few options I would like to see. If you happen to record a particularly long audio file it would be nice for the option to be able to play back the audio from certain parts of the text just in case you can’t remember what you said, in other words integrate some of our functionality functionality from the calibration window.
According to Nuance audio files can be up to 2 hours in length and like I mentioned earlier the quality will depend on your recording environment and just how well you trained DragonDictate 3.
I’m giving it 3/5 which may seem a little bit low despite a glowing review and how well its performed. The inability to re-train and calibrate the transcription process and lack of functionality means it does get marked down a little bit but by all means it’s still a great enhancements to the programme.