Your Mac may be your central workstation, but that doesn’t mean you want to sit at its keyboard all the time. With a digital voice recorder and MacSpeech’s iListen 1.5.2 — the only speech-recognition software that can transcribe audio files onto your Mac — you can dictate your memo, e-mail message, or short story on-the-go, and later download the file, push a button, and watch while iListen types what you said.
iListen ($99; with headset, $149; 816/373-4506, www .macspeech.com) isn’t perfect (after all, we did give it a mm rating, January 2003). But in our tests, the program was about 77 percent accurate when transcribing audio files into Microsoft Word. Three correct words out of four isn’t bad, and with a little effort you can raise that average and spend less time cleaning up errors.
Garbage In, Garbage Out
If you want speech-recognition software to understand what you say, you have to give it good material. This means having the right equipment optimized with the best settings.
Choosing a Recorder MacSpeech recommends only two digital voice recorders for use with iListen:
Olympus’s $150 DS-330 ( ; October 2002) — which we used for our tests — and $250 DS-2000 (800/622-6372, www .olympusamerica.com). Others may also work (they must record 16-bit mono files at 16kHz with no compression), but there’s no guarantee.
Headsets Help You can talk directly into your recorder, holding it like a microphone. If you choose this option, make sure the recorder’s Micsens (microphone sensitivity) switch is set to Dict. This lowers the sensitivity of the internal microphone and prevents it from picking up ambient sounds, such as a coworker’s laugh from down the hall. When you talk, hold the recorder 2 to 4 inches from your mouth.
However, speech-recognition software depends on clear, high-quality sound. Pops, feedback, and noise decrease accuracy. So does uneven volume — for instance, when you hold the recorder against your lips one moment and a foot away the next.
By using a headset, you’ll get an immediate boost in accuracy, as well as better mobility.
To work with a digital recorder, your headset must have a standard minijack plug, not a USB connector. (The VXi Parrot, which is available bundled with iListen, fits the bill.)
Nix the Noise Whether you use a headset or not, you should take advantage of the Olympus recorders’ built-in Noise Cancel feature, which reduces background noise. To activate it, select a file you’ve already recorded and press play. While the file is playing, press and hold the play button for a few seconds. The current Noise Cancel level should appear on the recorder’s screen. Tap the play button to scroll through the options until you reach the Hi setting, and then push the stop button. This is now your default recording setting.
Create a Profile for Your Recorder
To get the best results, you have to first train iListen to recognize your voice. You do this by creating a profile. Every user must have a unique iListen profile optimized for his or her speech — and each input device also needs its own profile. This is because your voice sounds different when it’s played by a recorder. Training iListen to understand the specific sound of the recorder is key to improved recognition.
The Hookup The first thing you’ll need to do is attach your recorder to your Mac. If your Mac has an audio-in jack, plug one end of the cable (included with the recorders) into that jack, and the other (L-shaped) end into the Ear jack on your recorder.
If your Mac doesn’t have an audio-in jack, you’ll need an adapter. Griffin Technology offers two options: the $35 iMic USB Audio Interface and the $100 PowerWave USB Audio Interface and Desktop Amplifier (615/399-7000, www.griffintechnology .com). In addition to acting as adapters, both devices boost the quality of your audio. In fact, I recommend using one even if you don’t need a USB audio adapter. (You already have the Griffin iMic if you bought the $149 iListen bundle.)
We were particularly impressed by the
PowerWave ( ; Mac Gems, June 2003). In our tests,
it significantly improved sound quality and as a result made training less arduous. Plus, it offers many other useful audio features, such as the abil-ity to connect stereo speakers to your Mac and to record and edit incoming audio.
Turn off your Mac before plugging the PowerWave into your Mac’s USB port. Use the cable to connect the adapter’s microphone jack and your recorder’s Ear jack. Flip the unlabeled black switch on the front of the iMic toward the microphone jack. Set the PowerWave’s Thru switch (on the back) to On and its Gain switch to Line. Plug your headset into the recorder’s microphone jack, and you’re ready to go.
Make a New Profile Before you start a new profile, make sure your System Preferences settings reflect your audio setup. Open System Preferences and choose the Sound pane. Under the Input tab, select the correct setting. For example, if you’re routing the recorder’s sound through a PowerWave, choose PowerWave USB Audio.
Open iListen and create a new profile (File: Profiles). The New Profile Assistant window leads you through the process of naming your profile and setting up your headset properly.
Trials and Audio Tests Once you’ve opened a new profile, iListen guides you through a series of sound tests. These help the program adjust basic settings to compensate for the quality of your recording device. Depending on your setup, this process may cause you considerable frustration. Speech-recognition software is picky, and record-ers don’t exactly boast the best audio quality.
To complete the tests, you need to dictate through your recorder into your Mac. Rather than holding down the recorder’s New button throughout all of the tests, use the more thumb-friendly option: simply press the Rec button once. Then delete the resulting file after you’re done.
To begin the Recording Volume test, click on the microphone picture and read the short passage on screen. After a few times through the sample text, iListen should indicate whether the sound quality is sufficient. If you find yourself reading the passage again and again with no feedback, it’s not sufficient. (We read the text 16 times on our first attempt before giving up.) In this case, adjust the volume on your recorder and repeat the test. Finding the perfect volume setting may take some trial and error. In our experience, volume level 9 was the magic number.
Still having problems? Set your recorder’s volume to 9, click on the Show Advanced Settings button, and tinker with the program’s Recording Volume manually. If you still have no luck, then break out the serious equipment: use a headset and consider springing for the Griffin PowerWave.
Once you pass the Recording Volume test, you can proceed with the Recognition And Silence Detection test. If this doesn’t go smoothly, open Advanced Settings again and try turning the Silence Detection level down.
Train, Train, Train! You’re now ready to begin the Learn My Voice training. This process helps iListen adjust to your voice and manner of speaking. You’ll start by reading a long passage of text as the program follows along.
If the program has trouble understanding you, try reading the whole phrase over instead of just the problematic word. To move past a particularly stubborn word, click on the Skip Word button. If you continue having trouble, try turning your recorder’s volume down slightly.
Once you’ve completed the initial reading, you’re ready to start transcribing. However, you can further improve iListen’s recognition by completing additional training exercises.
Polish Your Transcription Powers
Here are some additional tips that may help you get better results from iListen down the line:
Dictate Shorter Passages Rather than creating one hour-long file for an entire chapter of your novel, consider breaking the chapter into several brief dictation files. It’s easier to edit shorter files, since you can quickly fast-forward through the tape to listen to a particular passage.
Train for Special Vocabulary MacSpeech doesn’t yet offer special dictionaries for unusual vocabulary — for instance, legal, medical, or technical jargon. You can, however, use the program’s Learn My Writing Style feature to help it recognize the strange lingo you use most often. Create a text document that contains a list of specialized terms you commonly use. From iListen’s Speech menu, choose Learn My Writing Style, and then select your text document. This will spare you the work of correcting the terms again and again — for example, when iListen substitutes braid a cardigan for bradycardia.