59 replies to this topic

[Alek Ituin]

Kiiyor, on 14 December 2014 - 06:34 PM, said:

Well, except Google has one of the best voice recognition algorithms out there. The fact that people complain about it doesn't mean it's terrible - it means that there is no piece of software that can effectively translate lots of different languages, accents and intonations. Even Apple, having spent bajillions of dollars on Siri, have to allow the program to adapt and learn before it can consistently recognize speech from just one person.

If it worked, then heck yes, i'd be all over it.

Far from it actually. Google has one of the worst voice detection programs, ever. You can rig up a program that does better with a few coding classes and some clever code. And Siri? Pfffftttt, yes, it's "advanced". Actually, it's a half-arsed attempt at adaptive AI, with equally half-arsed voice detection. There are better voice detection programs out there, just like the Alien in Alien: Isolation is a better adaptive AI than Siri could ever hope to be. (It's also scary as f**k and good at dropping from vents when you're looking at panels that take hours to get out of. So there's also that.)

You seem to forget that mass-produced products, either physical or virtual, are not on par with the actual capabilities of a specific technology. It's adapted to ease of production, not being a top of the line tech demo. You can make Siri almost flawless, but she'd need a LOT more space on the drive, and a LOT more computing power to drive her. Which is why we don't have a nearly flawless Siri, it would make each phone cost upwards of 1000-2000 dollars and be the size of the classic "Brick".

For instance, we can make aircraft that go faster than Mach 5 (NASA X-43 for instance, clocked in at approximately 11'000kph), but you don't see them being used outside of testing because they're prohibitively expensive. What we use for air travel every day is pedestrian in comparison, because it's costly to use advanced technology on a large scale, and often requires a lengthy adaptation process that impacts efficiency.

Edited by Alek Ituin, 14 December 2014 - 07:05 PM.

[Andi Nagasia]

i dont agree,
1) Voip would be easier to implement, and easy to moderate(Mute Button)
2) a good talk to text is hard to make, and harder to implement,(does your Talk to Text always say what you mean?)
3) Pilot accents can cause problems with the Talk to Text system, and mistakes can easily be made,
4) not everyone can read text wail in combat, also when such text blends into the background,
5) text goes away after 5 seconds, dont notice it, o well your problem,

im not saying Talk to Text cant work, just that Voip is a better solution,

[Theodore42]

Having a hot mic on TS is one thing (what the OP is describing). VOIP in a game is totally different.

VOIP was first implemented in Counterstrike (maybe a few games before had it but nobody used it). Counterstrike was the first game that got people to go out and buy mics and the first game where the VOIP feature was commonly used. Even back then you had to play for hours before someone came along that needed a mute. And today there is even less of a problem. I have 660 hours in CS GO and I have literally NEVER had to mute someone. BTW, in csgo, the most annoying teammate is the one WITHOUT a mic, because he can't tell you what you need to know.

TS, where 12+ people have a hot mic and are opening bags of chips and eating and watching tv- is completely different from a push to talk VOIP feature.
And the tactical rewards far outweigh the worst case scenario. It is a game changer!

[FireDog]

The text chat widows rotates by too quick. At least in MW4 we could make it larger and scoll it back some.

[Kiiyor]

Alek Ituin, on 14 December 2014 - 06:55 PM, said:

Far from it actually. Google has one of the worst voice detection programs, ever. You can rig up a program that does better with a few coding classes and some clever code. And Siri? Pfffftttt, yes, it's "advanced". Actually, it's a half-arsed attempt at adaptive AI, with equally half-arsed voice detection. There are better voice detection programs out there, just like the Alien in Alien: Isolation is a better adaptive AI than Siri could ever hope to be. (It's also scary as f**k and good at dropping from vents when you're looking at panels that take hours to get out of. So there's also that.)

You seem to forget that mass-produced products, either physical or virtual, are not on par with the actual capabilities of a specific technology. It's adapted to ease of production, not being a top of the line tech demo. You can make Siri almost flawless, but she'd need a LOT more space on the drive, and a LOT more computing power to drive her. Which is why we don't have a nearly flawless Siri, it would make each phone cost upwards of 1000-2000 dollars and be the size of the classic "Brick".

For instance, we can make aircraft that go faster than Mach 5 (NASA X-43 for instance, clocked in at approximately 11'000kph), but you don't see them being used outside of testing because they're prohibitively expensive. What we use for air travel every day is pedestrian in comparison, because it's costly to use advanced technology on a large scale, and often requires a lengthy adaptation process that impacts efficiency.

Fun fact: most commercial and high end voice recognition software uses the Google Voice API, or some hacked up version of it. It's one of the reasons there was so much anger when Apple perma-banned Google voice recognition from their products (because it was flat out better than Siri at the time). Google actually uses the cloud to predict and analyze voice behavior, meaning that the strength of the voice recognition if you have an online connection can be far greater than what is stored on the phone.

Another fun fact: Siri's origins hail from DARPA.

You're right about specifically designed technology being better than generic commercial stuff, but why would PGI spend what would likely be an exorbitant amount of money on it?

[Argent Tnega]

Monkey Lover, on 14 December 2014 - 06:36 PM, said:

basic commands via keys and flares would work fine too.

Being attacked "flare location"
Attack here "flare location"
Group here "flare location"

This is in line with what I was thinking, Pings like League of Legends for instance. For more then just the Lance and Company Commanders.

[Captain Stiffy]

Argent Tnega, on 14 December 2014 - 07:53 PM, said:

This is in line with what I was thinking, Pings like League of Legends for instance. For more then just the Lance and Company Commanders.

I agree that this could also be a workable addition. I do also feel that the gameplay is often too complex for simple functions. But having a radial menu for commands would definitely be better than the absolutely nothing we have now.

[Alek Ituin]

Kiiyor, on 14 December 2014 - 07:48 PM, said:

Fun fact: most commercial and high end voice recognition software uses the Google Voice API, or some hacked up version of it. It's one of the reasons there was so much anger when Apple perma-banned Google voice recognition from their products (because it was flat out better than Siri at the time). Google actually uses the cloud to predict and analyze voice behavior, meaning that the strength of the voice recognition if you have an online connection can be far greater than what is stored on the phone.

Another fun fact: Siri's origins hail from DARPA.

You're right about specifically designed technology being better than generic commercial stuff, but why would PGI spend what would likely be an exorbitant amount of money on it?

I'm just pointing out that it's a possibility. I never said it was probable, but that it would be a viable idea with some utility to it. The difference is subtle yet critical.

But yeah, if Siri sprang from a DARPA project, it must have been an old one. Because there's several very accurate programs available for speech to text, all of which beat Siri at voice recognition. Obviously Siri has other utilities available to her, taking away focus from voice recognition, and she's based off of phone hardware. Kind of like a low-end VI...

Google is better off, but I didn't see much difference between that program and the speech-to-text on my phone when I used it. I'll give it another shot to see if they've improved the accuracy though... Admittedly, I didn't know they used the cloud, that knowledge does change things. Cloud based processing is a fantastic idea to boost available processing power, bravo to whomever thought to use it.


Interesting points you've brought up.

[FETTY WAP]

John Madden John Madden John Madden John Madden John Madden John Madden John Madden John Madden John Madden John Madden John Madden John Madden John Madden John Madden

[Kodyn]

Because reading tiny text on the side of the screen that disappears in an instant is better than what we have now?

No, not at all.

VOIP, or use TS, speech to text would be a massive waste of resources for a worthless feature.

[Haipyng]

VOIP is a nice idea except most game developers avoid it because in aggregate it's bandwidth intensive, surpassing the game bandwidth many times over.

Text to Speech is another nice idea. If you process the TTS on the server side it is less bandwidth intensive than plain VOIP since it is not being rebroadcast to multiple players, but it is CPU intensive and is going to need lots of hardware. Process the TTS on the client side and it will add load to the client side, something they have already backed away from due to poor machine performance on the users end. (lowest common denominator for performance in order to maximize subscribers).

Another issue is despite years of development, TTS is still not very good, especially with background noise. 75% of the time it would translating gibberish.

Edited by Haipyng, 15 December 2014 - 09:53 AM.

[Mechteric]

Sorry but current speech to text just isn't that great yet, its still prone to making errors where it really matters.

[Catra Lanis]

Text is good enough if people could bother to use it. I do agree that the text should stay a bit longer though or that we should be able to scroll back.

Edited by Catra Lanis, 15 December 2014 - 06:19 AM.

[Xyroc]

Captain Stiffy, on 14 December 2014 - 12:37 PM, said:

It needs speech to text.

eh ... some people hear better than they read others ignore text and need to be yelled at.

[Mechteric]

Catra Lanis, on 15 December 2014 - 06:18 AM, said:

Text is good enough if people could bother to use it. I do agree that the text should stay a bit longer though or that we should be able to scroll back.

The other problem with text is that while typing you aren't manning your controls. It can often lead to running into obstacles, not shooting at the enemy, etc. All around its not a great way to coordinate whatsoever.

[Desolator]

Captain Stiffy, on 14 December 2014 - 01:36 PM, said:

I might argue that listening to children, people belching, drunk idiots and control-freaks is even harder.

How is that any worse, than having to sift thru all the drunken grabble, childish banter,and control freaks filling the chat up with speech to text.

It isn't.

It is easy to tune out audio, it is impossible to scrolls thru all of the garbage that would fill the chat with speech to text, to look for actual useful information.

[Xyroc]

Desolator, on 15 December 2014 - 01:02 PM, said:

How is that any worse, than having to sift thru all the drunken grabble, childish banter,and control freaks filling the chat up with speech to text.

It isn't.

It is easy to tune out audio, it is impossible to scrolls thru all of the garbage that would fill the chat with speech to text, to look for actual useful information.

People make this silly argument every time. Like they never heard of mute before.

[AntiCitizenJuan]

It needs some form of fast, efficient mass team communication.
I dont care how it comes but it needs to happen.

I'd love to see Mech Commander be its own role and 13th slot for each team, with an RTS style view and fog of war. I doubt the engine is capable of doing that though. Would be nice not to force a commander to balance commanding all 3 lances and fighting himself

Edited by AntiCitizenJuan, 15 December 2014 - 01:18 PM.

[Captain Stiffy]

Desolator, on 15 December 2014 - 01:02 PM, said:

How is that any worse, than having to sift thru all the drunken grabble, childish banter,and control freaks filling the chat up with speech to text.

It isn't.

It is easy to tune out audio, it is impossible to scrolls thru all of the garbage that would fill the chat with speech to text, to look for actual useful information.

Maybe if you've never used IRC or only been on the internet for 15 minutes.

[Captain Stiffy]

So for anyone still doubting the technology... on my Samsung phone (which is yesterday's news now and is not awesome) I can dictate to it telling it to open apps or dictate text messages to it with a 99.9% success rate.