Move over, Siri. Intel is working with Nuance on its own voice assistant technology for Ultrabooks. Dubbed Dragon Assistant, the new software will come preloaded on select Ultrabooks sometime in 2013 and will allow users to conduct web searches, play music, check social networks, launch apps and much more just by speaking to their PCs.
We had a chance to go hands-on with a beta version of the software running on a Dell XPS 13 Ultrabook and came away impressed not only with its voice recognition capabilities, but with its friendly and attractive user interface as well.
Because Intel requires Dragon Assistant-ready Ultrabooks to have dual-array microphones (which must meet tight quality specifications), there’s no need to attach a bulky headset. Just talk directly to your PC and it talks back to you. Even better, the speech recognition dictionary is stored locally, not in the cloud. So, unlike Siri, which has to ping Apple’s servers to turn your speech into text every time you talk, Dragon Assistant can interpret your speech quickly, even if you’re not online.
In our brief experience, the software was highly accurate but, like all speech recognition applications, it didn’t interpret us correctly 100 percent of the time. For example, when we told Dragon to search YouTube for “Gagnam Style,” it gave us a list of videos for “numb style.” However, the software learns your speech patterns so an Intel rep who demoed the app for us received slightly more accurate results than we did. Also, if you are searching for something that’s stored locally like the name of a song, the software will be more accurate because it will only compare what you said against the list of song titles on your hard drive rather than the infinite possible queries you could perform on the web.
You can activate Dragon Assistant either by hitting the Ctrl key twice or by saying “Hello Dragon”. The latter method, of course, requires Dragon to keep your microphone on all the time just to see if you said the two magic words.
In our brief test, we appreciated the hands-free method of just speaking “hello dragon”, but during a conversation with another person, we accidentally launched the program. If you don’t like the idea of Dragon listening to you all day, you can disable voice activation in the program’s settings panel.
Once you’ve activated the program, it shows a tiny widget hovering on top of the left side of the screen. The basic widget just shows whatever Dragon is “saying” to you in text format and a sound level indicator that shows how well it is reading your voice. However, as you engage certain actions, a contextual display area appears below the tiny dialog indicator. So, if you say, “show my playlist” that area shows a list of songs in your current playlist and, if you say “Show Twitter” or “Show Facebook” it will show the four most recent updates on your feed with a scroll bar to see more.
If you have a scroll bar in the box, you can scroll up and down by saying “scroll down” or “scroll up,” though we didn’t see a command that lets you scroll all the way to the top or bottom of a list. To get a list of valid voice commands either for a particular function (ex: Facebook, Twitter, music) or for the app as a whole, you can simply ask “What can I say?” and a list of valid commands appears.
Dragon Assistant’s dialog box does not stay on top of the screen when it’s not needed. So, after you give a command and it waits a few seconds, the interface disappears only to show up again when you give another command or wake up the program.
Some types of voice commands put Dragon Assistant to sleep immediately upon completion, requiring you to wake it by saying “Hello Dragon” or double tapping the Ctrl key. Other types of commands keep Dragon Assistant awake because the software expects you to issue follow-up commands right away. So, if you’re performing a search and say “Search Amazon for Sunglasses,” Dragon gives you the result and then goes to sleep. However, if you ask to see your Twitter or Facebook feed, Dragon will remain awake, because you are likely to then say “scroll down” or “open the link in number two” as you stare at a list of updates.
One of Dragon Assistant’s most useful features is its ability to search using any of a half-dozen different search engines and shopping or media services. During our hands-on, we were able to search Google, Bing, Yahoo, Hulu, YouTube, Amazon and eBay by simply saying “Search [Engine] for [Query].”
So when we, for example, said “Search eBay for IBM Keyboards,” it spawned the default browser and pulled up a list of IBM keyboards for sale on the popular auction site. We didn’t test this feature personally, but you can also tweet the address of the web page you land on, making it easy to share what you’ve found without touching the keyboard.
One major feature of Dragon Assist is its ability to play music from your library on command. As soon as you put song files in Windows’ music library, Dragon knows about them and lets you call them by name. So, if you say “Play Smile” and there’s a song named “Smile,” Dragon will start playing it. If you want the music to stop at any time, you can simply tell Dragon to “pause.”
Even more interesting is Dragon Assistant’s ability to play music by genre. When we said “play rock,” it began playing all of the songs in our library with meta data marking them as “rock.” If you say “Show playlist” while playing a genre, Dragon will display a scrollable list of all your available songs in that genre. So when we asked it to show the playlist of rock songs, we saw two songs and asked it to then play “We are the Champions”, which it did.
Intel told us that Dragon also supports video play, but the demo system we were using did not have any local videos stored in its library.
Dragon Assistant features deep Twitter and Facebook integration that not only allows you to post and check updates, but also, as stated above, shows you a list of your friends’ latest updates as part of its widget. In our testing, we asked Dragon to “show Twitter updates” and it responded by displaying the latest four messages from our feed with numbers next to them and a scroll bar that let us scroll down (by saying “scroll down”) to view more.
When we wanted to click a link in a tweet, we simply said “Show the link in number 3,” where the number 3 was the third tweet on our list. However, we had to be careful what we said, because when we said “Show number 3,” Dragon pointed our browser to the permalink for the Tweet itself, not the link within it.
Facebook worked exactly the same way as Twitter, showing four messages with a scroll bar and numbers next to each message that let us call them by number. When we said “show us number 3,” on the Facebook list, it sent our browser directly to the Facebook post associated with that number.
Though we didn’t try it, both Facebook and Twitter allow you to ask for a list of your private messages, not just your feed. You can compose Tweets and Facebook messages just as easily.
Unfortunately, Dragon Assistant does not support any other social networks at this time and an Intel rep said there were no immediate plans to add Google+ or other options. The program does support Skype dialing by voice, but our demo system did not have Skype installed.
When it launches, Dragon Assistant will let you dictate text into any Windows application you want. However, the beta version only works with certain programs. In our testing, we were able to voice type into web fields such as the search box on Yahoo. With our cursor in the text box of our choice, we held down the Ctrl button (as opposed to tapping it twice) and waited for a microphone icon to appear in the middle of the screen. We then said “Laptop Magazine” and watched as those words appeared in the text box.
According to Intel, you can only talk for up to 30 seconds each time so don’t plan to recite Hamlet’s soliloquy without pausing every half minute to let Dragon catch up. We think this dictation feature will be even more compelling when you can use it Microsoft Office and other productivity apps.
Dragon can also launch applications for you if you just say “Open [Name of App].” So when we said “Open Paint,” the program launched Windows Paint. However, in the beta version, you must say the exact name of the program with no variation allowed. So, if the program is named “Adobe Photoshop” and you just say “Photoshop” it may not work.
Dragon provides a fairly detailed settings panel that lets you control a number of options in Dragon Assistant. Using the panel, you can set whether or not Dragon wakes on voice or Ctrl button only, configure Dragon to work with your calendar, POP email, Twitter and Facebook accounts and even set periodic voice updates that proactively tell you that you’ve gotten new messages.
Unfortunately, because Windows 8 does not have the proper voice control APIs available yet, Dragon Assistant runs as a desktop app. It cannot control Windows 8-style (aka Metro) apps and, when you are in the Windows 8 UI rather than desktop, you won’t see Dragon’s UI appear on top of your work. However, Dragon does continue to run in the background so you can give it voice commands and get its response; you just can’t see its messages unless you’re in desktop mode.
Though iOS and Android have their own voice command systems, it’s particularly compelling to be able to command your PC with voice. Intel told us that the functionality we’ve seen is just the beginning and that the app will eventually support a whole range of new commands including natural language queries. We can’t wait to see where this goes.