Jump to content
Sign in to follow this  
SaSSharp

AI Voices

Recommended Posts

So when I was watching game play videos of Arma 3 when I heard the AI's voices I was shocked, they are the same horrendous voice acting found in Arma 2! please someone tell me they have confirmed the adding of smoother more realistic sounding voices.

Share this post


Link to post
Share on other sites

All I gotta say is... remember, it isn't even in beta phase yet. So what they show isn't close to a finished product. I'm 90% sure the voices will change lol Not to sound like an ass

Share this post


Link to post
Share on other sites

Because they haven't made voices yet so they used the same ones as in Arma 2

Share this post


Link to post
Share on other sites

There is no way to convincingly voice act the literally millions of lines the AI has to say. I hope they don't waste resources on it. Read the subtitles or play MP like everyone else.

Share this post


Link to post
Share on other sites

It is perfectly possible to have very nice voices, the recipe is very simple: few changes in dubbing configs, profesional voice actors and lots of time to do lots of work.

But my personal opinion is that the chance of that happening is very slim.

The voices were screaming for help in OFP (but we didn't care because OFP was awesome), then they were horrible in the XBox game, after that they were terrible in Arma 1, not much better in Arma 2, and then again in Arrowhead.

So, obviously the voices are not considered important enough to be made better.

Share this post


Link to post
Share on other sites
It is perfectly possible to have very nice voices, the recipe is very simple: few changes in dubbing configs, profesional voice actors and lots of time to do lots of work.

So how will you say a word that can be used in 20 contexts with universal intonation that will fit everywhere?

It's quite tricky actually. When you start talking you breathe in and breathe out upon ending a sentence or a bit of it so whatever was in between has the same tone. So if you will say Alpha Three Five Six it will come out natural. But it's really tricky when you record Alpha, Three, Five, Six separately.

Share this post


Link to post
Share on other sites

Then you say the word 20-ish times :)

I dont think there are too many different contexts nor intonations.

Basically you can have four contexts:

1.) first word in a sentence

2.) word in the middle of a sentence

3.) last word of a sentence

4.) single word

...and maybe three intonations:

1.) agitated

2.) informative

3.) quiet

There would be a lot of additional sound editing work, but its doable.

Its only a matter of available resources (time/people/money), the question is: would it be worth the resources you put into it? I dont think so, and i dont mean quality-wise.

Share this post


Link to post
Share on other sites
please someone tell me they have confirmed the adding of smoother more realistic sounding voices.

Yes, Bohemia have found a new voice actor -

CcSyGCKIkJ4&feature=related

Share this post


Link to post
Share on other sites

The voices are not from voice actors, they're not all from real people.

If you were to make every sound file ever to be used, it'd take a year more to come out and the game would be 5gb more of data.

Share this post


Link to post
Share on other sites
I dont think there are too many different contexts nor intonations.

Numbers are used everywhere in the game. So are unit type names. So are distances.

Basically you can have four contexts:

1.) first word in a sentence

2.) word in the middle of a sentence

3.) last word of a sentence

4.) single word

...and maybe three intonations:

1.) agitated

2.) informative

3.) quiet

In ArmA2 you have 1 intonation and it does not help. Because the word "Three" is recorded separately from "Move", "To that house", "10 O'Clock".

It only seems to you it's doable with magical editing. Falcon 4 also has a similar system where sentences are generated and it also sounds robotic. It's a bigger problem that can't simply be solved by throwing hours into it.

The voices are not from voice actors, they're not all from real people.

They are from real people actually. In ArmA2 one of voice sets for AI was coming from BIS beloved voice actor they put into every game since OFP Elite (the dude voicing main character in Harvest Red and the Delta Force dude in OA). It's too bad they've ditched that voice set in OA.

Edited by metalcraze

Share this post


Link to post
Share on other sites
The voices are not from voice actors, they're not all from real people.

If you were to make every sound file ever to be used, it'd take a year more to come out and the game would be 5gb more of data.

As far as Arma 2 and OA is concerned, i am confident they are from voice actors, real people.

---------- Post added at 11:24 ---------- Previous post was at 11:19 ----------

But i am not saying it would not sound robotic, just that it would be 90% better than the current state.

The fact you have 1 intonation in Arma2 doesnt help not because it cannot be done, but because they didnt put enough effort into it - they just didnt do it, but not because its not possible.

---------- Post added at 11:28 ---------- Previous post was at 11:24 ----------

Anyway, you could allways record samples for almost whole sentences like "...move to that tree", "...move to that car", "...move to that building", etc., and then add only "one"/"two"/"three" to the start (there is a tiny pause anyway, so it would fit ok).

Lots of work, but perfectly doable.

Share this post


Link to post
Share on other sites

I'm ok with the voices in Arma 2 (Maybe I got used to it) but I wonder if it's better to have more generic sentences recorded in one block (as "move to that point" or "move over there") to sound natural, or more flexible and fragmented samples to allow to cover all situations (get cover / behind that / tree / at / 6 O clock).

I actually think it'll be best to have good sounding voices saying generic stuff like "get over there" complemented by a more complete text radio chatter reading "get over here to that blue house at 6 O'clock etc.."

This way you could record multiple variations of this kind of generic orders with different styles to avoid the repetition, and show anger, fear etc... it would fit the AI well too during battle.

Share this post


Link to post
Share on other sites

[/color]Anyway, you could allways record samples for almost whole sentences like "...move to that tree", "...move to that car", "...move to that building", etc., and then add only "one"/"two"/"three" to the start (there is a tiny pause anyway, so it would fit ok).

Lots of work, but perfectly doable.

There's also move to that rifleman, move to that AT gunner, plane, tank, APC, bush, forest etc. There's also distance. X O'Clock is added too. It still will sound out of place. Oh and if you want to record all that too as one sentence - prepare for 100 GBs of voice recordings alone.

Maybe far in the future when a terabyte game will be a norm and BIS will be a megacorporation with unlimited time and resources you will have a unique phrase for everything.

As it is right now the only other option is Dragon Rising where everything was 100m north even if it was 200 or 500 or 1000 to 10 oclock. It gave you absolutely no clue about where the enemy is but at least it was a single sentence!

Edited by metalcraze

Share this post


Link to post
Share on other sites

I don't mind the robotic procedural voices, at least they're useful all the time. Recorded voice lines are usually only useful in specific, sometimes unique, circumstances. As such they're a poor use of resources IMO.

One thing I might ask - PLEASE make sure the voice actor who says the word "close" pronounces it "close" (e.g. that dog is too close) and not "cloze" (e.g. please close that door) :) always irritates me in ArmA2 ;)

Share this post


Link to post
Share on other sites
There's also move to that rifleman, move to that AT gunner, plane, tank, APC, bush, forest etc. There's also distance. X O'Clock is added too. It still will sound out of place. Oh and if you want to record all that too as one sentence - prepare for 100 GBs of voice recordings alone.

Yes, i know there are more variants, but lets do some simple math here:

For each actor and language...

The basic default radio protocol consists of approx. 87MB in 850 sound samples.

Substract 55MB which takes the samples for SOM module, Warfare, core, formations, and combat - which we either dont need, or they already are as a whole sentence.

Substract 2MB for various other samples which are already also recorded as a single sentence (but i was too lazy to cherrypick them and count them).

Add basic stealth radio protocol, which consists of 29MB in 441 sound samples (its significantly smaller than the default, because many sentences or words are not used in stealth communication).

So, you get 59MB for one actor, lets say each sample would have to be recorded 10 times (which is surely more than needed), you get 590MB for one voice (but more realistic would be something like 300MB).

If we use 10 actors (which is again more than needed), we get 5.9 GB (or 3GB if we consider those 300MB per actor).

That is perfectly ok, and FAR AWAY from yours "100 GBs of voice recordings alone".

Share this post


Link to post
Share on other sites
Yes, i know there are more variants, but lets do some simple math here:

For each actor and language...

The basic default radio protocol consists of approx. 87MB in 850 sound samples.

Substract 55MB which takes the samples for SOM module, Warfare, core, formations, and combat - which we either dont need, or they already are as a whole sentence.

Substract 2MB for various other samples which are already also recorded as a single sentence (but i was too lazy to cherrypick them and count them).

Add basic stealth radio protocol, which consists of 29MB in 441 sound samples (its significantly smaller than the default, because many sentences or words are not used in stealth communication).

So, you get 59MB for one actor, lets say each sample would have to be recorded 10 times (which is surely more than needed), you get 590MB for one voice (but more realistic would be something like 300MB).

If we use 10 actors (which is again more than needed), we get 5.9 GB (or 3GB if we consider those 300MB per actor).

That is perfectly ok, and FAR AWAY from yours "100 GBs of voice recordings alone".

5,9 GB is not perfectly ok, the game has to fi on a DVD. Also you forgot a very important thing: Making this would cost money, propably a lot of money, and time. Valuable resources that can be used to make something more usefull.

Share this post


Link to post
Share on other sites
5,9 GB is not perfectly ok, the game has to fi on a DVD. Also you forgot a very important thing: Making this would cost money, propably a lot of money, and time. Valuable resources that can be used to make something more usefull.

I agree with Tonci87,

5.9GB for voices is rediculous. The time and money it'd take isn't worth it.

If you don't like the voices, there's a slider in the audio settings to turn them off.

Share this post


Link to post
Share on other sites
5,9 GB is not perfectly ok, the game has to fi on a DVD. Also you forgot a very important thing: Making this would cost money, propably a lot of money, and time. Valuable resources that can be used to make something more usefull.

It has to fit on a DVD? WHY? there are games that are on 2 or even more DVDs, and nobody have any problem with that.

Today those few more GBs are nothing.

And yes, it would cost resources - that is perfectly obvoius right? and its exactly why i said i dont see that happening,

but not because its not possible/doable, but because they wouldnt want to put those few extra $ into it as there is NO real NEED.

It would be nice to have, but its not needed, hence it wont happen.

Share this post


Link to post
Share on other sites

They are adding new voices, for opfor atleast. Because they currently have no Persian voices in the game. No, Takistan does not speak Farsi they speak Dari.

Share this post


Link to post
Share on other sites

I'm not bothered by the voices in the game. I'd prefer they spent their time on AI, or animations, or the physics engine. The voices do what they need to do, which is transfer orders and information in a succinct manner. I'd much rather hear "2, Taking Fire!" than "I'm getting f***ed in the ass over here!" (a la BF3).

Edited by Snowden
typo

Share this post


Link to post
Share on other sites
All I gotta say is... remember, it isn't even in beta phase yet. So what they show isn't close to a finished product. I'm 90% sure the voices will change lol Not to sound like an ass

Thats what they said about Armed Assault and Arma2 before release. Don't get your hopes up.

If BIS would let you simply rearrange units in the command bar I would be happy. Then they could simply have the officer say: "1 through 4, Go to, that House" instead of:

1, Go to, that House.

2, Go to, that House.

3, Go to, that House.

4, Go to, that House.

Only then to have you change your mind and make them go somewhere else, and then imagine having 10 units. AI move faster individually rather than in a standard formation.

Edited by Victor

Share this post


Link to post
Share on other sites

You can already divide your squad/platoon into teams in Arma 2. Then instead of "1,2,3,4 go to that house" the leader says "Team Red, go to that house" or "Team Blue, go to that car." Or did you mean AI leaders?

Share this post


Link to post
Share on other sites

Theres a thread deep in the BIS vaults from Arma2 that debated this to death, and same points come again. This isnt a scripted game with a narrow option list, so no matter what way you swing it, voices just cant be as fluent as we would all like with the multitude of options, resources taken to cover it, time taken with everything else that needs to be done, and as mentioned bloating install with such audio.

Best voices so far (IMO) were BAF DLC which just made it a little more fluent, but you will never get much more than that.

Then you say the word 20-ish times :)

I dont think there are too many different contexts nor intonations.

Basically you can have four contexts:

1.) first word in a sentence

2.) word in the middle of a sentence

3.) last word of a sentence

4.) single word

...and maybe three intonations:

1.) agitated

2.) informative

3.) quiet

There would be a lot of additional sound editing work, but its doable.

Its only a matter of available resources (time/people/money), the question is: would it be worth the resources you put into it? I dont think so, and i dont mean quality-wise.

When the game is released I hope to see your AI Voice addon pack, because only a dedicated team could do it how we would like it and BIS have enough to debug the game before release and everything else that is needed :)

Share this post


Link to post
Share on other sites

Sorry, no voice pack from me, i am already busy enough regarding BIS games.

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
Sign in to follow this  

×