When will games start having better voice acting?

Ther are exceptions to this of course, like Mafia 2, but in general? Its a freaking joke how terrible and horrific voice acting can be in games, so damn cheesy and cliche, its importance is way too underrated, it seems like game companies squeeze it in at last minute after a quick recording session.

Voice acting can honestly make or break a game

 

just cause 2..... that games voice acting is all you need

 

Uncharted 2 and Killzone 2/3 have AMAZING voice acting...

Not to mention Half-life 2... And Portal... Portal, oh Portal...

 

Well, voice acting is time consuming, and expensive. Yes, good voice acting is a nice touch. But I'm willing to give mediocre voice-acting a lot of leniency, if it means that the developer focused their time / resources on great gameplay and creating a polished finished game. To me, a fun, bug-free game is far more important than voice acting.

 

It'd be nice to see a major game developer come out with a new game sometime soon that comes out with great gameplay, a polished/finished product (meaning no freaking bugs that should've been fixed in the beta stage, etc etc etc, or at least just a FEW bugs), or something like that....

*waits for it*
*waits for it*
*waits for it*

.....well that was fun...

 

The last 3 new-release games I bought were Dead Space 2, Monday Night Combat, and Fallout New Vegas. Never had any problems with those. I also bought Civilization 5 and Dead Rising 2 at release, also no problems on my end.

In fact, the last game that I bought which had any noticeable bugs / problems was Singularity back in July 2010. And I've bought 22 games (both old and new releases) between then and now, all without problems. And even the problem I had in Singularity just magically "went away" because Steam's auto-patching system eventually patched my problem out.

Maybe I'm extremely lucky, or maybe you're extremely unlucky. But I've been pretty happy with the quality of game releases in the past few years.

 

bad voice acting is better than no voice acting...i hate games where you have to read everything. I refuse to play them because i am a slow reader and its boring. But yes better voice acting would be nice. Also more games need to have voice acting. Reading everything is dumb

 

I wish some games with voice acting would let you turn it off because I read much faster then they talk. I hate waiting for characters to finish a sentence or paragraph that I've already read 3 times.

 

all games i haev played allow you to click to continue as they talk so you can skip it. Also i find voice acting gets me into the world more. It ameks me feel like i am there

 

You're extremely lucky. Fallout: New Vegas, in particular, is a notoriously buggy game. Quests can tangle themselves very badly and start locking things up and can only be unraveled through the console. Weird things can happen. It was a great game nonetheless, it just didn't feel finished because of all the bugs in it on release. I haven't played enough of Dead Space 2 yet to make a judgement, but it seems ok so far.

I was being overly judgemental on purpose though, because it does seem a trend of certain studios, some of them very major ones, to release some very buggy software only to patch it a few days (rare) up to several years or never later. Most of them that stick out in my mind are the ones built on the engine that Oblivion uses (namely FO3 and FO:NV) which have bugs in the engine that have been around since Oblivion came out and have not been remedied since.

 

games will have better voice acting as soon as it becomes lucrative for good voice actors to pursue video games as a primary source of income/buzz.

As it stands right now, i think there's no question that a voice actor could make 10x + more money in a pixar flik than anything expanding out into video game territory.

 

I'm willing to forgive Bethesda for Oblivion having some of the worst voice acting I've ever heard (I mean, one male and one female voice for each race, some of which overlapped...really?) simply because of Patrick Stewart being in it for all of 10 minutes.

Also, Final Fantasy X. I'll grant some lenience since it was the first game in the series to have voice, but...god damn, try a little harder.

 

Mark Hamill doing the Joker in Batman: Arkham Asylum is as good as its gets; which is no surprise because he already does the voice of joker in the Batman animated series. Also, the voice acting in Magicka is awesome!

Voice acting seems to be a mixed bag, most the games have a few actors who are fantastic, and then a few that are mediocre-terrible. I definitely think it does add something to the game. Dragon Age is a good example, a few of the characters were fantastic, and lots of em were terribad.

 

There's quite a few, including of course the half-life series, but most recently, Amnesia: The Dark Descent:

><param name="movie" value="http://www.youtube.com/v/MKpbcJVpl1I?fs=1&amp;hl=en_US&amp;rel=0"></param><param name="allowFullScreen" value="true"></param><param name="allowscriptaccess" value="always"></param><embed src="http://www.youtube.com/v/MKpbcJVpl1I?fs=1&amp;hl=en_US&amp;rel=0" type="application/x-shockwave-flash" allowscriptaccess="always" allowfullscreen="true" width='480' height="390"></embed></object>

 

I still think they need to come up with a speech acceleration engine. I brought this up before. Once you get enough voice samples you could set a programmer to modify all the voices by using a slick interface and make as many voices as you'd like, all computer created. Be cheaper to pay a programmer or two than a dozen different quality voice actors.

 

thats awesome ^^

 

try assassins creed series and metro 2033. they had me at their accent effects.

 

My name... is Bolo Santosi!

That's because AC (2 and Brotherhood, at least) and Metro 2033 had Italian and Russian actors, respectively, doing most of the voice work. In the case of AC the Italian actors coached the non-Italians on how to affect an Italian accent.

Me, I kinda love bad voice acting. Makes games hilarious. Like the RE series.

 

Have you heard the Japanese version? The voice acting in that game outclasses the American voiceovers by FAR. Very, very, far.

But then again, Japanese voice actors are usually highly skilled, and highly paid too.

 

Play Mass Effect 1 and 2. Superb voice acting!

 

That problem is actually much more complex than you have made it out to be. We don't even have great synthetic instrumental tones, and those are EXTREMELY simple straightforward compared to voice.

Consider even the piano, where you just have a finite number of keys that can be pressed with a varying force, released, and 3 pedals. It's easy to imagine a simple mapping to recorded samples, but that vision is naive. The problem of recreating that sound from prerecorded samples is extremely difficult. This is because the strings in the piano interact with each other, and that interaction varies for every combination of keys, force applied, and time (as in the sequence of key presses, a combination keys could be pressed at the same time, or in any sequence). Every combination generates a unique sound. The worst part is that we are extremely sensitive to these variations in sound. Anyone can tell the difference between the a digital piano sample that came from a record map, and an actual piano recording.

The situation on other string instruments where the musician interacts with the strings directly are significantly worse. In addition to the complexity that must be maintained with string interactions, you now have to consider a much wider range of inputs. With the piano, there are just 88 keys and 3 pedals. All of the inputs are restricted to pressing and releasing the keys and pedals. How many things can you do to a guitar to make sound from the strings, and how many variations can you do of these things? It's a tremendous task just to pin that down, and then for each of those you have to record all the instrument interactions for that sound, for enough pitches to mask the fact that what we are doing isn't continuous.

Just as the guitar and violin are much more intricate to recreate than the piano, the human voice is extremely more intricate than than the guitar and violin. Without going into it too much more, the number of sounds we can make are quite large.

The technology today is at the point where we *can* recreate the piano pretty well, but not from pre-recorded samples at all. Rather, the only strategy that produces good results are done from a system-modeling approach, where you have a fast computer simulate a physical piano in space and calculate the sound results of the inputs (key pressing with varying force, releasing, pedals) dynamically while playing. There are way too many sounds to pre-render them all and store them in a sound bank.

We have basically no progress on robust digital recreation of the guitar or violin. Again, the human voice is extremely complicated compared to these.

---

Also, for a variety of reasons, we cannot just generate the desired wave-form for the desired human sounds manually (similar to the process of 3d modeling and animation, applied to human voice)

The biggest reason is that animation is done on a 30-60 frame / second field, and we can generate our animation by specifying about 1 model state per second or even less, and interpolating the rest. 60 frames per second, and about 1 key animation frame per second can be sufficient for most applications.

With sound, we are much more sensitive. Sound is CD and mp3's, etc is done on a 44,000-48,000 frame / second field. The human voice usually exists in the 200-7000hz range. We have extreme acuity to be able to detect changes in this domain.

I know we talked about this before, but the fact is this technology is no where close to being available. Progressive research on speech synthesis is still focused on making speech understandable, and producing organic sounding speech. But you have to understand that these requirements are not even close to those for voice acting - where the goal is not just understandability and organic sound, but recreation of character. It's a huge leap forward. We need to be able to at least model the voice to be adaptable to style, different types of voices, different intonations of sounds, etc etc etc before we can even consider generating a tool to put it all together.

TLDR version: It's frustrating how much you underestimate the undertaking you recommend. Also, I think you want an acting-quality speech synthesis engine, not an acceleration engine. It needs to work before it needs to work faster. Last, this technology is not even close to existence. There are many prerequisite technologies that we just don't have, and won't have, for many years.

 

But we lived with blocky graphics, which was no where near anything we'd consider real fidelity. Why not start somewhere with voice? I mean video cards require more power than any other component and can run more expensive than the rest of the PC components combined for high end cards. We need to work towards that with audio as well.

 

At this point, I think I'd need a list of modern games with bad VA.

 

you're confusing things here. we already have great audio fidelity. we also have great voice acting (not across the board or anything, but we don't have great animation across the board either). also - even animation is progressing away from the direction you are talking about.

model quality, environments, resolutions, and textures are all getting better. animations are getting better too. one of the big things pushing animation quality forward in games is mo-cap. it's the voice acting equivalent of animation, and it's the technology that's being used more and more in AAA game titles that include great voice acting as well.

 

There is just way more involved in simulating speech than you think. They can't even make it organic sounding yet (which is several orders of magnitude more simple than being able to procedurally generate character voices). That would be a big step up from where we are now, but only one of many steps required to get to anywhere that looks like what you are talking about. I don't know what else to tell you, except that even huge landmarks that need to be crossed before it's even a possibility are still nowhere in sight, and if you are waiting, don't hold your breath. Maybe you'll see something like this in your lifetime if you aren't too old currently.

The first big landmark is being able to automate speech at all. We made it that far, at least. But, CG speech is still very broken and mechanical.

The next step will be the automatic generation of speech that at least sounds organic. This is a lot better, but still no where close to where we need to be. It's the difference between being able to make a sound that sounds like it came from a guitar versus being able to recreate more or less any sound that a guitar could make (and having enough knowledge to be able to put something together with it, which is also very complicated).

Then, you'll start to see automated voices in our day to day lives. Not prerecorded voices (as in the annoying phone menus), we have that already. You'd need CG voice for this application. We're looking more for something like: you would be at the grocery store and there would be a touch panel that goes across the shelf. If you touch the panel, a CG voice could speak to you and tell you basic information about any item on the shelf, in any language. That sort of thing. That will happen way before we get CG voice acting. It's still hugely more difficult to get voice acting than it would be to get the scenario as described above. It would render the speech dynamically from text, and it would sound human to you, but not expressive. Until we can get to the point where the grocery store voice that tells you information about the product sounds completely indistinguishable from a human voice, we can't even get to the next step (do you not see how sci-fi this is for us today? we are a such a long way off from here, even)

The next big landmark would be the abandonment of instrumental recording of music in AAA movie titles. Consider a movie like Inception or even Dark Knight. Movies with that type of soundtrack would have to ditch the orchestra for a programmer and CG instruments long before we could do the same with voice.

After all that, it will be worth talking about. If we have these things in order then we would have the prerequisite technology in place to start tackling CG voice acting.

 

simulated voice will bring forth the "uncanny valley" for speech synthesis.

it already sounds goofy as it is, and high quality professional speech synthesizers are just too weird sounding to put into high production games. it works well for "robot" characters perhaps, but human characters would be too disturbing.

anyways, DCUO has some good voice acting.

 

My point is that you have to start somewhere. If they initially decided against making 3D graphics because it didn't look "real" then we'd be nowhere. You don't get anywhere by doing nothing, you have to start somewhere. We have the voice synthesis the way it is because it hasn't been pushed in any mass consumer avenue at all. That's how commercialism works. Not a bunch of "it won't work" but "how can we make it work".

Start small, with certain aspects of a game, then build on it from there. Gaming is much more tolerant of poor acting or overall presentation than movies. Look at the trash we buy and keep buying. So it'd make sense to start with gaming. Sure it may take some programmers with a tool to tune the voice, but it'd be a lot cheaper and easier to implement thousands of lines of text than have a dozen or more voice actors. I'm not saying straight text to speech, you can't. There's inflection and emphasis where you need it not to mention dialect and purpose mispronunciations (Put the em-phass-is on the sy-laaa-ble).

 

On a recent episode of PCGamer Podcast, they actually talked about how the more life-like that games tried to be, the more our brains would reject it. IT's because our brains are so wired to picking up something "wrong" with an image / sound that is supposed to be human. That is why voice-acting and character-acting in games like Oblivion trigger a red flag in our heads.

When you look at things that intentionally do not try to be life-like, our brains totally accept it. Examples are:
* - Zombies. Nobody ever says "the zombies don't look right in Game_XYZ"
* - Stylized art. Our brains accept the worlds created in Borderlands, World of Warcraft, Team Fortress 2, etc, because they aren't attempting to emulate reality.
* - Stylized voice. GlaDOS. Sim-lish. They sound acceptably funny, because they are supposed to sound funny.

But re-skinning Half-Life-Alyx into Adriana Lima is a nice touch.

 

There is so much more to speech than proper pronunciation. I hope you can appreciate how vastly fluid and organic and complicated speech is... I'm beginning to wonder if we aren't operating on a mutual level of understanding.

I agree that it would be great to have text and maybe corresponding emotive clues and have the computer just take care of the rest and hear a person talking expressively and organically on the other end.

As stated above, there are a variety of landmarks we haven't even approached that we will need to pass before we can even begin to tackle the problem of text driven character speech.

If you are really interested in pushing the field forward, start tackling those problems.

You could also look at it this way.

Just by taking a wave form and editing it manually, we can create whatever sounds we want. We could manually create the wave form of a conversation if we wanted. It would take an astronomical amount of time.

We could reduce the problem by being able to stay within the proper sound set. To do this, we will need to be able to model and simulate the entire vocal system using physics. That would be the lungs, the vocal chords, and the mouth (tongue, teeth, lips). That would be a good first step. Even if you can't model these things perfectly, iterations of the model driven approach is really the only way to ever generate the proper sound set dynamically. You would be able to change variables in your simulated system to create sounds from different types of people and with variation in emotion. At this point, we could start animating human sounds much the way 3d animation is done (well, mostly the way 3d animation was done, we're moving fast towards the equivalent of voice acting in 3d animation, because it produces better results). Unfortunately, because the human ear is much more picky for sound than for vision, our work will still be astronomical. It will take expert voice animators who dedicate themselves to the art of making good voices (voice programmers instead of voice actors) about 100 times the amount of time it takes the animators to do their job, because they will need at least 100 times as many key frames and the art of it is going to be somewhat more complex than the visual animation.

To reduce the problem further, we can have an AI help us make the key frames. The more intelligent our AI is, the more we can rely on it to make keyframes for us. It would have to be extremely advanced compared to anything we have today to be able to contribute key frames at all, and it would pretty much need to generate all of the key frames to be usable. To do this, the AI will have comprehensive information of all aspects of vocal language, written language, word meaning, sentence formation, word interaction, colloquialism, inflection, intonation, phrase formation, word interaction, our physical voice model, and all the interactions that exist between these domains. If we could get to that point, then we could be at a place where we could program out language instructions that resembles our current language filled with hints to the AI and have it go the rest of the way. Would it be as good as a voice actor? It certainly could be if every aspect of the system was well done and the AI was advanced enough to simulate creativity on it's own (the programmer instructions consisting of words and tonal hints won't be close to enough on their own to make it sound even decent)

That's it. Simple. Solved. That's how it's going to happen and we will have it in about 30 years or so.

I still don't understand why you seem so anxious about it. It's not even close.

Again, think about the difference between recreating a sound that sounds like a guitar, versus being able to simulate all of the sounds that a guitar could make, and then having an AI play *that* simulated guitar for you on it's own in an intelligible way that sounds good to you. Think about how much more complicated the second task is compared to the first task. Consider that we barely have the first task done for the piano, haven't finished for the guitar, and we are just beginning with voice.

 

htwingnut:

how far do you think we are from being able to generate organic sounding speech (not mechanical, ie. fluid)?

how far do you think we are from being able to generate speech that sounds human? (ie. it won't bother people who hear the sound coming out of a human character model OR indistinguishable from a human)

how far do you think we are from being able to generate human speech that can at least capture a range of emotions suitable for a primitive character?

how far do you think we are from being able to generate dynamic character speech?

how far do you think we are from being able to generate dynamic character speech that can approach the quality of good voice acting?

----

I admire your resolve and go-get-em-ness, but I wish you would understand how your envisioned solution has overlooked a wealth of non-trivial problems. They can be solved, and they will be solved, but there is no reason to rush the solution without having a relatively complete understanding of the intermediary steps. The only way to get from point A to point B is via a path from A to B, and that path has quite a few steps along the way. About 30 years of steps. It just makes more sense to focus on relevant research.

----

I'll give you a real world example of what is frustrating about this, and why I am writing so much. It might help you understand better where I am coming from and why a "let's tackle character speech simulation now!" attitude bothers me.

A civil engineer came to my school a while ago and gave a lecture about adapting the human circulatory system for self-repairing buildings. He was a civil engineer. He was literally doing research on recreating the human circulatory system at the nano-material level for use in buildings so that they could be self repairing, like the human body. Again, this was a civil engineer. He didn't have any specialty in mechanical engineering, or chemical, or biology. Nevertheless, he presented all this research to us as if it were relevant.

Is this enough of an analogy to see the problem? There is so much other research and so many other problems that need to be solved that are prerequisite to self-repairing circulatory system based buildings that it doesn't make any sense to talk about it beyond conjecture. At the moment, the same is true about speech simulation. Start solving the problem for string instruments or join up with Piano-teq to improve the piano model if you are interested in working on the vocalization aspect of speech, or you can work on formal linguistics, which we will also need a better understanding of in order to make this work.

 

the human voice has more character to it than the human [physical] model in my opinion.

you can move human models easily because it's a matter of moving bones in the models (you can use mocap to reconstruct human behavior in that sense). and since the behavior is much more generalized than speech, then it can be applied to all models without the need for too much variation. blinking is blinking. running is running. you can't generalize sound as easily since everyone sounds slightly different from eachother. it isn't a matter of altering static variables. the computer would have to dynamically change the variables since people generally don't even sound the same saying the same word over and over and over again.

 

Just installed and playing Chronicles of Riddick : Assault on Dark Athena and have to say it has some of the best voice acting I've heard yet !

But, I also believe that if it's a choice between great voice acting and mediocre gameplay , budget-wise, I'd prefer the money be spent on gameplay !

 

If you have the physics model and enough time, you can sequence together speech. Even without the model, if you have enough time you can still sequence together speech. However, because of the level of detail we need to tolerate artificial sound, and especially human speech, the amount of time it would take to sequence it manually, model or not, would be astronomical. We would absolutely need automated intelligence behind the model to handle basically all of the lifting for us. Keep in mind that the more we have to contribute to the system, the less this whole thing makes sense and the more voice acting makes sense. If it's considerably more complicated than text -> speech, we might as well just get a voice actor. If I have to specify the variation of intonation over time, and specify stresses, volume, and other parameters, it's going to take longer to do all of that than to just have a voice actor.

You need to be able to pass text to a program that can interpret the dialogue, possibly given some hints about the mood or context, and then determine how to render the dialogue based on the context, what is being said, the emotions of the characters, other similar high level parameters. If you have to specify low level details about the speech, it's going to take you longer than a voice actor.

 

Games need to have voice actors with memorable, distinct, and unique voices. Sadly, not a lot do.

 

Generally alot of the problem is budget, and society. No one takes pride in their work anymore, and it shows. But there are tons of great examples of voice acting in games, heavenly sword, MGS series, Half life 2, so on and so on. Another problem voice acting has to over come to be believable is lip sync. Alot of games (even more so for import and translated games) have a poor time syncing the lips to the voice, and as a human this is a key element to believability. But if L.A. Noire has anything to say about it (pun intended) then we should be starting a golden age of voice acting and believability. There is no reason games can't be motion captured to precision anymore, heck you can do motion capture with a web cam now.

 

The classic Half-Life had the most memorable characters EVER! I will never forget the scientist's voices! I am sure other games have good voice acting, but this is the only one I have personally played all the way through.

 

In the MMO genre, GW2 sounds like it'll have some decent voice acting going. Not the most fantastic in a video game, but for the genre, pretty damn decent.

Legend and Legacy – The Norn in Guild Wars 2 – ArenaNet Blog

Against the Wall: Humanity in Guild Wars 2 – ArenaNet Blog

 

In the interest of saving masterchef from having to write another wall of text...

Just imagine your video game voiceovers done by Microsoft Sam or Stephen Hawking. That's pretty much the point we're getting at.

 

^ Lol. Very good analogy and valid point.

 

Like Daranik said above, LA Noire seems like it's bringing a whole new level of voice and game acting. Every time i see it, it looks more like a movie in video game format.
Looks like we're moving completely away from the old days of video game characters having mouths that either dont move or simply open and close without any real connection to the words they're saying.

 

yeah.

just to sum up my wall-o-texts:

- that is where we are today in terms of synthetic speech
- having a system where you define a character by his speech style, then type in text and mood hints and maybe a few intonation hints and have it produce character speech that wouldn't be extremely jarring to hear (again imagine someone speaking to you as stephen hawking or microsoft sam, etc.) is MANY years away (20 years)
- anything short of this technology and it will be cheaper to hire a voice actor to act than to hire a programmer to animate a voice
- big studios producing AAA games will still use voice actors as long as the results are significantly better, which will still occur for several years after such technology becomes available (at least 5 more years)

 

And companies would actually have to want to invest some money in this sort of project. All programmers, whether at Intel, EA, or Valve, need their morning donuts.

 

I hope so , I was watching a conference on making convincing characters and rememberable characters and alot of it had to do with motion capturing and how far its come. Like LA Noire shows you don't even need a matrix dot grid anymore, they simply make a poly mesh for 30fps, then they turn it into a texture map and apply it to a character. The manurism and small qwarks in the characters faces are remarkable. Heavy Rain was great and was heading in that direction too. Think if something like Skyrim or Metal Gear Solid or mass effect used a facial technology like that one in LA Noire, we'd be all blown away by the performances, maybe even bring tears.

Im slowly delving into game development and pondor these questions often, how can I make a player angry at what happens in the game, or how can I make them cry, or happy about the outcome. I think alot will have to do with how convincing the characters are.