Poll: How accurate is XB1 voice control?

[Shifty Geezer]

For owners of XB1, how accurate do you place Kinect's voice control in terms of accurate response to instructions? That is, if you say 5 instructions including one repeated because it's missed the first time, that'd be 4 hits out of 5 commands or 80%. I want an estimate for your current usage now you've gotten used to it, with perhaps a post for how long you've been using voice control and whether you've noticed it improve and your use adapt to it. It'd also be good to get usage observations, like does the environment change and does that affect accuracy?

 

[BRiT]

For basic controls it's 100%.

For trying to pick out channel names on a complex name out of 340+ channels I'd say it's around 95%. It all depends on the channel name. For instance the channel FXNOW it wants the command as "xbox, watch F-X-N-O-W", spelled out instead of F-X-NOW. Once you know this quirk then I can get hit rates of 100%. Naturally I have not tried to tune to every specific channel, so there may be other ones that might need tuning as well. Given the channels I do watch, it's exceptionally accurate.

I've been using Kinect voice commands since launch weekend (Sunday November 24th), around 3 hours each weekday evening and around 10 hours a day on the weekends.

The environment is very noisy when the heat kicks on at various cycle points in the day. I'm also using it while I have music playing from other sources. It has no issues in detecting my commands.

The only issue I have is when paging through the guide, it's listening times out before I'm ready to issue the command. I'm not quick enough to tell it "PAGE DOWN" so when that happens, I simply say "xbox, select" and "xbox, page down" and it's onto the next page. When it's still listening I merely indicate "page down" or "later" and it goes to next page or moves to later time.

I don't know if it tuned itself to me or if I tuned myself to it.

 

[Cyan]

Sigh. I had to re-write my entire post from scratch, I was writing it in a different thread and something went wrong I don't know what I did.

Kinect works almost flawlessly for me -I am not even English native, it's not my mother's tongue- when it CAN actually listen to me.

I have the Kinect placed very close to the console -subject to change- and the slight noise coming from the Blu-Ray spinning might affect speech recognition.

It's odd because sometimes I just whisper to it and it works like a charm. Yet other times I have to increase the tone of my voice for it to listen to me.

When it listens to me -which is most of the time- speech recognition works well like 90% or more of the time for me.

Example "Xbox Select" -an essential voice command right out of the box I learnt recently-, or Xbox Record That, etc etc.

It mixes up Xbox Snap and Xbox Unsnap sometimes, at least for me, and it's all in the timing. Xbox Unsnap always works, surprisingly enough, Xbox Snap requires a more precise timing and pronunciation so Kinect doesn't confuse it with "Xbox Unsnap". I basically say "Xbox sNap" -with an almost unpronounced s-.

When I say Xbox Snap and then Games DVR -it works 99% of the time, it recognised it as Video instead of Games DVR once, I wonder why (perhaps I pressed the A button and don't remember/realise?-, the Start New Clip, the Start Recording, then Cancel or Stop Recording, all of these work without a flaw.

FIFA 14 also works so well especially when you take into account you can say actual players' names, so does Need for Speed Rivals. It seems like EA
games work very well with speech recognition. Need for Speed Rivals is awesome in that sense, I have never missed a single voice command. FIFA 14 is more or less the same, except for players with a strange name or a name I don't know how it is pronounced.

Commands like Xbox Go To Forza Motorsport 5 or Xbox Go To Killer Instinct, Xbox Go To Powerstar Golf, Xbox Go To Ryse Son of Rome, Xbox Go To NBA 2k14, Xbox Go To Crimson Dragon, Xbox Go To Upload, Xbox Go To My Games & Apps, Xbox Go To Settings... etc etc, work flawlessly in my case.

Xbox On works like 50% of the time. Yesterday I was drinking a glass of milk and walking towards my laptop turning my back to the console and since I was drinking my voice wasn't totally clear and barely whispered Xbox On like 2 meters away from it without even looking at the console and I saw the Xbox logo lights turning on. I was like.... "WTH". What did I do for it to work so well?".

To add to that, the Xbox On voice command doesn't work in other languages other than English, afaik.

I set the console language to English, with a very high rate -I insist, non native, Galician is my maternal language- although I also tried Spanish, which I speak perfectly, and I mean that, 100% knowledge of it, so to say. In Spanish the rate is even better for me, it basically never fails.

I think the reason is that it is a language which is *poor* at vowels. A sounds like the a from Sam, E sounds like the in term, I sounds like the i in Tim, O sounds like the o in Oz, and u sounds like the in impromptu.

The experience has been very positive for me overall. I'd chose 75-80% because of the fact that it doesn't seem to hear me all the time in some specific situations. Were I to choose one of your options based on speech recognition I'd go with your second option.

It still needs some work though. I have friends whose sons or daughters go crazy over voice commands and they work so well, and cases like this one where speech recognition is flawless for one member of the family but not for the rest.

http://forums.xbox.com/xbox_forums/xbox_support/xbox_one_support/f/4275/t/1633528.aspx

 

[AlphaWolf]

I put 90%, but for the base UI controls it's very close to 100% for me. The browser gives me some issues (still above 75%) with show address bar.

 

[NRP]

Same here. UI controls seem like 100%. Some of the TV and browsing stuff is a bit less, so I chose 90%. The least reliable voice command is "Xbox on", for some strange reason.

 

[dagamer]

I've owned my Xbox One since launch (November 22nd).

On average, commands work about 80% of the time, but some commands are skewing the average. I'll put them into 3 general groups:

Near perfect: "Xbox Go to *app*", "Xbox Go to *game*", "Xbox Go Home", "Xbox Pause", "Xbox Play". This is also work when a lot of audio is coming out of my speakers during a show/game.

Usable: Most app specific commands, especially when it's not generic text like "Item 1" and stuff.

Needs work: Pre-update, "Xbox On" worked 25% of the time, now it's 50% for me. This command doesn't work well from my couch when all other commands do, I need to be 2-3 ft away from Kinect for it to work reliably.

One nice thing is I have yet to have the Xbox trigger commands from a show I'm watching. My 360 Kinect was so terrible at one point I had to unplug it. Not so on the One.

 

[eastmen]

I have a speech impediment which affects TH , CH , SH sounds mostly. Sometimes the Kinect has trouble understanding me with that stuff. But i'd say its about 90% of the time for me. The quicker I try to say something the less it works. When I take my time saying things it works a 100% of the time.

 

[Jogi]

Just horrible experience.. mayby it is me, "air conditioner", plasma tv buzz, sitting too far away, moon phase, or something.. it would be nice it worked better, some voice control features are "ok".

 

[LightHeaven]

For every voice control it's something above 90%... I've got some wrong matches when trying to do a Bing search (but so far very few), and that was it. To say it's was not 100%, the go to <app> command has taken me to the wrong app once or twice... But in my case, it's damn near perfect.

If I use in Portuguese, of course, in English, that's not my primary language my success rate is a little lower, but mostly due me not saying the right command.

Oh, and at least in Portuguese almost all voice commands are there. The only exceptions are Xbox on, and the TV related commands (volume up/down, go to channel, etc, which don't matter much because there's nothing supported that uses hdmi in yet)

 

[Silent_Buddha]

I would have put 100% except that it occasionally will miss something. I'd say it's probably around 98-99% for me. Similar for my friend that I bought the console from.

Then again, I'm in the demographic that is probably the easiest to get voice recognition correct. Virtually no accent of any kind. It's why English call support centers were so popular in my area until companies found it cheaper to outsource them to other countries.

I've had 100% recognition for the basic commands. Xbox On, Xbox Turn Off, etc. The only times those have failed is when I didn't say the correct command. Like at the beginning when I tried to do "Xbox Off" instead of "Xbox Turn Off". Which then made me mess up "Xbox On" as I was saying the incorrect phrase "Xbox Turn On", which obviously doesn't work.

Regards,
SB

 

[Arwin]

For Xbox Snap, could it be that you can just say the same thing to switch it on and off? That would make sense to me.

You could theoretically make this topic also for the PS4, by the way, although the voice commands are currently very limited so that it wouldn't really be a fair comparison. For instance right now there's no way to control a DVD menu with voice commands, but up to that point it works well, as does starting any games. It makes me appreciate voice commands a bit more - it clearly makes things easier than finding the remote and even remembering where the right buttons are for that particular remote layout (we got a new tv recently)

 

[dobwal]

Around 95% most of the time once Kinect trained me on some of its commands. LOL.

 

[Scott_Arm]

I said 90%, though for me it might be higher, excluding "Xbox On" which is at best two out of three for me. I don't think it adapts at all, in terms of learning your voice. My room is very quiet, without any noticeable background noise. There are no fans, or noisy fridges nearby. It seems to deal with my speakers quite well, which are two or three feet to either side of the Kinect. I have noticed some noises, like the loud high-pitched squeak my old chair makes when it rocks, can disrupt voice commands. Background noise could be an issue. So far I haven't tested voice commands from another room to see how it'll work. Within my living room seems to be good, even if at an odd angle. The most important factor in success rate is the cadence of speech, adding a pause between Xbox and the following command. I don't seem to have to raise my voice. Speaking in a low tone and at a normal volume will work fine.

Oh yeah, I noticed volume up and down are not as reliable for me. The pause, play, stop, rewind, fast forward etc all seem to be normal, but volume up and down seem to have a lower success rate for me. Must be the way I'm saying volume.

 

[dobwal]

I said 90%, though for me it might be higher, excluding "Xbox On" which is at best two out of three for me. I don't think it adapts at all, in terms of learning your voice. My room is very quiet, without any noticeable background noise. There are no fans, or noisy fridges nearby. It seems to deal with my speakers quite well, which are two or three feet to either side of the Kinect. I have noticed some noises, like the loud high-pitched squeak my old chair makes when it rocks, can disrupt voice commands. Background noise could be an issue. So far I haven't tested voice commands from another room to see how it'll work. Within my living room seems to be good, even if at an odd angle. The most important factor in success rate is the cadence of speech, adding a pause between Xbox and the following command. I don't seem to have to raise my voice. Speaking in a low tone and at a normal volume will work fine.

I wonder if the reason for that is that "Xbox On" lacks access to the cloud solution. I wonder how Kinect VC would work with no internet access.

 

[Scott_Arm]

I wonder if the reason for that is that "Xbox On" lacks access to the cloud solution. I wonder how Kinect VC would work with no internet access.

It doesn't use the cloud for any of the voice control, unless you do a bing search. I was pretty impressed by the bing results. It could pick up words like angioplasty without issue.

 

[Deleted member 11852]

I've still not laid hands on an Xbox One but this is pretty much what I expected based on using voice recognition technology for about a decade.

What does surprise me is the acceptance of imperfection in the implementation. How many of us would long tolerate a remote control or controller that demonstrated a mere 5% failure rate?

Very interesting. For what it's worth, it the voice stuff is a learning implementation, it will get better. The trick with these things is keep people to use it enough that occasional mistakes become so rare you rarely remember them.

 

[Shifty Geezer]

What does surprise me is the acceptance of imperfection in the implementation. How many of us would long tolerate a remote control or controller that demonstrated a mere 5% failure rate?

Two points. 1) I already tolerate remote controllers that aren't 100%. Sometimes you have to wiggle them around to get them to register, often when the batteries are getting low. 2) The 5% failure is around fringe tasks and not the major functions which most are calling pretty much 100%. So in general use it works, and only in some aspects where the names get complicated does it fail, which is still better in some cases than trying to type in on a virtual keyboard. Texting on a phone can be 80% or worse at times, but we all still put up with it. For finding games and content, using the name is very intuitive and likely faster than various button interfaces even if you have to try a couple of times.

 

[AlphaWolf]

What does surprise me is the acceptance of imperfection in the implementation. How many of us would long tolerate a remote control or controller that demonstrated a mere 5% failure rate?

1. My remote isn't 100% either, it will often miss a button push (pushing buttons too fast or slow will screw it up) taking me to the wrong channel.

2. Kinect performs complex actions that would take multiple button presses and you can recover from a failure faster than if you had been using a controller in the first place. (going from watching TV to playing a specific app could be a half dozen button presses without xbox one (change inputs, navigate to the app and launch vs 'xbox goto appname').

3. I don't have to look for kinect before I can issue commands.

 

[Deleted member 11852]

What device do you own where the remote control isn't reliable? Low batteries excepted, that isn't normal and stuff like that drives me nuts. NUTS! But I admit to have a low tolerance of things not responding.

No problems with texts, though. Practise makes perfect!

EDIT: I do recall a JVC DVD player that used a tiny remote powered by one of those circulate watch batteries. I tolerated it because it played back DivX from CDs or DVDs, but yeah - the range on that things was pretty poor.

 

[Scott_Arm]

...
2. Kinect performs complex actions that would take multiple button presses and you can recover from a failure faster than if you had been using a controller in the first place. (going from watching TV to playing a specific app could be a half dozen button presses without xbox one (change inputs, navigate to the app and launch vs 'xbox goto appname').

...

This. Voice control is just faster. Flipping back and forth between the video I'm watching and a game is one voice command away. If it fails once out of ten attempts it's still faster than navigating menus. Unless I am playing Battlefield I do not touch my controller at all.