Up close with Mountain Lion: Dictation
Though it falls short of full Siri integration, Mountain Lion's system-wide Dictation tool does bring iOS's transcription functionality to the Mac.
Anywhere you can type text on your Mac, Mountain Lion lets you dictate it, too. On iOS, the onscreen virtual keyboard provides a little microphone key that you can use to trigger Dictation mode. Because your Mac can't update your hardware keyboard dynamically, Mountain Lion instead requires you to use a keyboard shortcut instead.
By default, that shortcut is tapping the Fn key twice. (You can also use the new Start Dictation item near the bottom of the Edit menu.) But you can customize the keyboard shortcut in the Dictation & Speech preference pane in System Preferences. (Before Mountain Lion, that pane was merely called Speech.)
On the Dictation tab, you can turn that functionality On or Off, tweak its shortcut, choose the microphone, and specify the language you'll be speaking in. Dictation needs to know that last bit before you start talking; it will obviously impact the way the software transcribes your words. Supported languages include English (in U.S., UK, and Australian variants), French, German, and Japanese.
When you adjust the keyboard shortcut, Apple includes a few suggestions of its own: In addition to the default Fn Fn option, you can choose to use double-presses of the left Command key, right Command key, or either Command key. If you prefer to select some other shortcut, you can—but you can only create traditional keyboard shortcuts (such as Command-Shift-Option-D), not any based on double keystrokes.
Speak, don't type
Whatever shortcut you choose, Dictation is available anywhere you'd normally type: writing text, filling in Web forms, composing emails, and so on. You'll see a microphone on your screen near the insertion point; the mic lights up to reflect your current volume.
Mountain Lion wants Dictation to capture your every word as precisely as possible, so it will automatically pause iTunes and stop your laptop's fans to create as quiet an environment as it can. Both will start up again if appropriate when you're finished dictating.
Dictation doesn't transcribe word by word as you speak; rather, it waits until you're all done. When you are finished talking, you can press your Dictation keyboard shortcut again (just once, if it's a double-keypress shortcut), press Return, or click the Done button by the Dictation microphone. If you instead press Space or another key, however, any memory of what you said is lost faster than you can say, "Whoops."
As Mountain Lion processes your spoken words, you'll see some animated purple dots. The moment your transcription is ready, those dots get replaced by your dictated text.
Your Mac doesn't actually do the heavy duty of converting your speech into text; that's instead handled by Apple servers elsewhere. For that reason, Dictation works only when you have an active Internet connection. Apple also requests your permission to upload certain bits of your data to its servers—notably, the first and last names of the folks in your Mac's Contacts—so that it can better transcribe what you say. Without my address book, for example, Dictation is inclined to transcribe my last name as
Freedman, and my colleague Dan Moren's last name as
Dictation and punctuation
As in iOS, Dictation in Mountain Lion supports a variety of specific vocal cues. For one thing, you'll need to get comfortable speaking your punctuation—the software can't infer where your em dashes go on its own. Instead, comma, get used to speaking like this period.
You can use numerous other verbal shortcuts, too. Common phrases to commit to memory include
new line, and
quote ... unquote. You can also speak the names of various symbols, such as
frowny face. In our early testing, Mountain Lion's Dictation doesn't support some other cues that iOS's version does, including
all caps on or
all caps off or
cap for capitalizing individual words.
While you don't need to take time to train Dictation manually, Apple says that it can improve its accuracy in transcribing your particular voice over time. If Dictation isn't certain that it properly transcribed specific words or phrases, it will underline them in blue; Control- or right-click on them to see alternate suggestions. Even if Dictation doesn't underline other text in blue, it may well have made transcription errors; you'll always want to read back over your transcriptions carefully.
In general, Dictation performs quite capably, but is far from flawless. While you can speak naturally, you'll get better results from Dictation if you're careful to enunciate, particularly on potentially homophonic words and phrases. Dictation is patient: You can take brief pauses to gather your thoughts and the service will keep listening to you, without shutting itself off. If you leave Dictation open for very long time, it will eventually stop itself and transcribe everything it heard. But in general, you have a good long time to speak, and pauses aren't a problem. It never hurts to take your time.
If you aren't yet a regular dictator (in the sense of someone who regularly speaks to a machine, not a despot), using Dictation can feel a bit awkward at first. If you stick with it, though, it can become a powerful way to gather thoughts and write text of varying lengths, when using your keyboard would be difficult or unwieldy. And it works just like typing—if you highlight text before you start dictating, the transcription will replace your selection; if you move the insertion point elsewhere in your document before you dictate, that's where your transcribed text will appear.