Originally Posted by hstroph
I would've enjoyed hearing about your assessment of Utter! from the Play Store (currently free) ... the reviews are generally great.
Utter! Voice Commands (Beta)
(Please skip reading this if you don't want all the gory details of my initial impressions, and experiences trying to get Utter started).
First thing: it's asking for root permissions. Unusual for such an app, and I'm not sure why. Got a notification saying "Utter has become self aware".
I'm in its SETTINGS now, and I see it has "native recognition language". That option allows you to choose from every country going. Haven't seen that one elsewhere. I've read lots of counts of VC apps not recognizing regional dialects of English. I am assuming this is an attempt to correct that, because I'm already assuming the app can't understand other languages than English. It has a lot of other settings you don't normally find elsewhere, so this is an ambitious release. Unfortunately, most don't do anything yet (e.g. Shake to Wake, Wave To Wake, Interaction Level, Wake-up Phrase, etc.) Pressing "Shake To Wake" just makes the phone vibrate, but does not turn the Red X to a green checkmark).
"Utter!" bills itself as being "different", in that it claims it is "super functional", "fast" and can be activated from the background. All great goals, but not exactly unique. "Vlingo" and Maluuba are super functional and fast, and I believe Assistant and Eva and probably Jeannie can work in the background. Although I've not seen any other that (potentially!) has as many options of waking the voice app as this one does. I like that it comes with a command list in the start screen, so you don't have to guess what commands you need to say to make it work.
But as far as making it work... well, it's not immediately obvious to a newcomer (who didn't watch the video or anything...) how you actually start the thing. There's an "ON" button and its green, but I'm talkin', and it's not squawkin'. There's a blue "Utter!" icon to the left of that, but pressing it doesn't do anything. Guess it looks like I'm gonna have to click on "user guide" to get this puppy in gear! .....WHICH takes me to a thread on the XDA forum! (Ok, THAT's gotta change!).
It starts off telling me it has created an icon on my launcher. Great. Except I don't have a launcher, per se. (I'm running a WP7 clone, which has no dock). I don't have "screens", either. It can't create an icon for me anyway, all of my icons are custom. Fortunately, I can create a widget and fortunately again, I did see it in my widget list (that's not always the case).
Clicking on the icon gives a short vibration, and it says in the notification bar that it's listening. I enter a command to see if it is; "What's the time?". It responds in the notification bar with "Computing". When I hit the start screen widget (you can also activate Utter in the notification bar, or in the case of the Desire, the hardware search button), I get a (female) voice, saying "Can I help you?". And "What's the time" get's the actual time!
Even though my first command was simply to ask for the time, not a difficult task for any Voice Control app, I was pretty impressed with the way Utter! worked, right off the bat. Because I did not try to speak in a slow cadence, or particularly clearly. I spoke as I would normally speak that request, and the response I got back was quick, and in a pretty natural female voice, with just a slight roboticish tinge. It spoke of the app's potential. True to its copy, Utter! is so meant to be used in the background, it's not at all clear to me right now, how to get back into the app (to check the command list)! Going through it's widget, the notification bar, or the search button, just gets the voice control activated. I had to go back in via the "OPEN" button on the Play store page!
UTTER! AT WORK
For a meaningful comparison, I will run Utter through the same tests I ran on most of the other vc apps:
Q. "Send an email to Robert"
A. "That command meant nothing to me".
I actually like this response! No nonsense. RIGHT OFF! (rip). I knew I might be tripping it, because the command in the command list just says "Email (contact)". So this time...
Q. "Email Robert".
A. "Please tell me what you want to say to Robert Watt"
(I told it "Please let me know when you will be coming tomorrow from New York")
Once Utter registered a too long pause in my dialogue (1-2s), it ended that and asked:
Q. "And the subject?
To which I responded "Hello what's up".
Then Utter asked me "Do you want to proofred it?"
I said "yes", and the email client list pops up, allowing me to select which email client I wanted.
Once in the email client, I saw it had chosen one of the two email addresses in my contact file (but with a comma after it). But it chose the main
one I prefer to use, so that was great.
This is the subject as Utter! wrote it:
"Hello what's up".
This is the body of text that it showed me:
"Please let me know when you will be coming tomorrow from New York"
I wanted to say more and got cut off, so it's not very long. But in both subject and body, Utter hit the accuracy at 100%
! It even put an apostrophe in "what's"! (It could very well have written "watts"!). It adds "Sent by voice from Utter!" as your sig. Pretty cool! Although I haven't done it, I would say this thing is ready to be put to a head-to-head challenge with Siri, for voice-activated email messaging.
Q. "Text Robert"
As with email, it asked
A. "Do you want to proofread it?".
Brings up choice of SMS clients, but no matter which I pick, the message I spoke isn't passed to the client. I don't know if it works should I say "no" the proofreading option. (I don't want to risk it, because I don't know if Utter warns you before sending an SMS. And I got hit with over $30 in SMS fees from my carrier, just from testing apps recently!).
Where is the nearest gas station"
Utter! has a lot of a commands, that it can understand. A LOT
. But "Where is",
isn't one of them. So I couldn't ask it to find a local business; ie. a currency exchange counter. It will navigate to that business (opening up Google Nav), presuming I knew the location.
It does however have "SetCPU to +governor".
I could see where that could come in handy. If I'm kidnapped by Mexican drug lords, and they've bound my wrists and feet, and my Android phone is in my fanny pack.... I could shake my bum-bum, wake up Utter!, and then she could change the governor setting on my device by remote control, so that it wouldn't risk overheating on a hot Mexican day. Yeah, who needs to know where a gas station is, anyway? It certainly wouldn't help one bit in overclocking my HTC.
To which it specifies home/mobile,
and you say "Home"
and it passes the number to your dialer. So yeah, that's about as good as it gets! (If not as pretty). However, unlike Siri and some of the VC apps I tested, there is no way to ask it to dial a direct no. This shouldn't be too hard to implement in a later version.
Q. "Skype (contact)"
I'm not sure Siri can make direct calls through Skype. But I tried this, since I have it, and it does work. The name of my contact is "A.", which it understood once I said "A period". (Only, the call goes to "null" instead of the contact I named; but that may be a technical problem not related to the app).
Open "File Explorer"
It had no problem opening the app, and dare I say it, was probably the fastest of them doing it. This trick alone might be worth the memory it consumes remaining active at all times! It could mean no more having to get out of an app and go hunting for apps to open on your phone.
"Set an appointment for Dec 31st at 7:45PM".
Utter responds "There you go", and shows a written confirmation of the event on the screen (from 7:45p to 8:45p), using the stock calendar app. Nice.
: "Set an alarm for 8:45"
This is where I ran into some of Utter's limitations. Realizing that it's still rigid when it comes to understanding natural language. It responds more to set commands, and if you stray too far from what it knows.... it tells you it couldn't decipher the command. Now in the command list, it says you set an alarm in hours and minutes. So I didn't like having to think how many hours ahead do I want to wake up or be reminded of something? But ok, I spoke in the language I thought it wanted to hear:
"Set an alarm for 12 hours and 10 minutes"
Which got me a confirmation in a small box upper left of the screen that showed a text version of what I spoke. But then another smaller box at the bottom of the screen said it didn't understand that?! I then realized the larger box can be pulled open to reveal a list of what I might have said, repeated a dozen times. You can press on each one but it doesn't do anything. In such cases, you really have to be precise to get the right response:
"Set an alarm in 12 hours and 10 minutes"
That got "Ok, it's done".
But how to confirm? It did not bring up the stock alarm app, so I'm not sure where it is set, as there is no "Show alarms" command. (Hint to self: it's in the stock alarm app)
: "New note"
This could prove to be very useful, taking voice dictation. I'm not sure what other voice apps will pass your speech to a note app. But it appears to require "Evernote" app to work, with either "voice note"
"Play knockin' on heaven's door".
Now, that's a song title. I didn't even try to be clear or slow. It said "Playing closest match" (if you don't have the song, it will play whatever sounds closest to what you said!). But it got the song right and played it directly in the stock player.
Translate to French:
"The rain in spain falls mainly on the plane".
I got a quick vocal response, but I couldn't make out a word of what was said. It sounded like some alien language I'm unfamiliar with. Then again, I was on Pico TTS at this point, so that might not help. Still, kudos for trying. Siri (on the 4s) can't even go that far. Would have been nice to see support of a text printout of the translation.
"What's the weather tomorrow?"
A. "Outside it's 1 degrees centigrade. That's 31 degrees Farenheit. And clear.
Again, quick, sharp responses. But "Utter!" is very vocal oriented, and goes without the pretty visuals and in-depth reports (on weather, etc.) you get from, say, Maluuba or Siri. Asking for the weather tomorrow or this week will get you the same response (the weather of today
n.b. Like all other vc programs I tested, it can't understand the province of "Nayarit" in Mexico. (Damn good thing I don't live there, is all i can say). But at least it was different in its wrongful guesses ("Norwich Mexico, Night Rich Mexico, 9 Rich Mexico, Nitric Mexico"). So they weren't kidding when they said it was different from Siri and other Siri clones. I take that to imply it's not searching Wolfram Alpha for these answers.
Q. "What's the capitol of Norway"?
A. "That command meant nothing to me, I'm afraid".
(Mind you, it printed the command in the on-screen dialog box, showing it understood what I said. Tapping the selection in that box yields nothing).
Again, Utter! is showing its tendency to be inflexible in your interactions with it. You can't just ask it anything your mind comes up with, as you might do with a more "Assistant-oriented" app. But... if we try "Search (Google) for the capitol of Norway", we get a Google page with a map of Oslo on it. It's kind of like going into town on a donkey trudging over rocky cliffs, while your friend named "Siri" goes there in a red Porsche with two hot chicks riding shotgun. But.. .you both still end up in the same place, more or less.
Q. Spell "supercalifragilisticexpialidocious".
A. "Really sorry, but I didn't understand that request".
Now, here I was trying to trip up the engine, and said the word real fast. Yeah, like a smartass. So I'm not surprised it didn't respond in the positive. But I am surprised that it spelled the word perfectly anyway, if you looked at the confirmation clipboard! (which had multiple variations of the spelling, and no. 2 was correct). This scenario was repeated exactly the same way, when I asked it to spell "heterogeneous".
Q. "When was Bill Clinton born?"
A. "Sorry, but I don't think I heard you on that command".
Yet here again, it printed exactly what I said to its clipboard! So it did hear me.
Utter! has a lot of commands (some of which NO other vc app has... or perhaps wants to have?). But its responses seem limited to those commands. Obviously, "when was"
is not one of them, so I couldn't make a comparison here. You can come close by saying
Q. "Search Google for Jacques Derrida".
To which it responds "Searching Google for shock give me to...
"..... (The first Google hit is "Why does my car door give me electric shocks?
" That's GOT to be more interesting than deconstructionism any day...).
Q. "Define (word) "sundry".
A. "Sorry, but this feature is currently disabled. "
I guess that's why they're calling it a 'beta'. But, it did understand the phrase correctly, as shown in the clipboard.
Though I wouldn't list it as my favorite (because for one, there's some important things it can't do), "Utter!" is a very promising voice control app. It seems to be oriented towards Android geeks. (ie. Judging by the "take a screenshot", "reboot bootloader", "run tasker" commands...). It does many things other such apps don't even think of, some of which can even be useful to some. For example, you can toggle wifi or bluetooth by voice, ask it where your car is, browse files by voice, read the clipboard, change the volume, translate phrases to various languages, check your battery stats and probably change more system settings than you knew you had.
It has a ways to go, before it's suited and ready for work. I think those who like Jeannie would probably like Utter! a lot too. It goes where Jeannie fears to tread... (right into your root files...).
Runs in the background, can be activated a variety of ways, including a hardware button. This alone turns it from a party novelty to something that, potentially anyway, can be used to change the way you interact with your phone, and make things a lot more productive. Assuming you always have an internet connection going.
Although you can speak to it at a fluid pace, it's not big on artificial intelligence. Skewing more towards having the capability to do many things that can be triggered in few ways, rather than a few things that can be accomplished in a larger variety of ways.
: After a little more fooling around with Utter, I came to realize the "Shake to Wake" does function correctly, even though the red x did not immediately turn to a checkmark. Not even sure how the "Wave to Wake" works yet, but it did say "enabled"....