122: Face Job

[Sam Sundberg]

Face Job

"'The actors are our vehicles for emotion. We need to make sure that we capture every last bit of their performance and translate the emotion into the game. You know, getting 80 percent of that is quite easy. But after that every additional percent takes a lot of effort,' Cage says. 'Right now, we're working on lips and the revealing of teeth. It's incredibly complex, a matter of fractions of a millimeter, and if it's not right you'll immediately notice. We're also working on the tongue. You may not think you see my tongue moving while I'm talking, but believe me, if it stopped you would be horrified.'"

Sam Sundberg speaks to Quantic Dream's David Cage about Heavy Rain and the power of the face.


Permalink

 

[AK-00]

If they can make scenes from daily life interesting to play, they'll be wizards, not animators. That sort of thing isn't even interesting to *live*.

 

[Katana314]

True, but I think actually hearing a believable conversation could really feel like a big change from the normal. Games have always tried to establish moody characters that have inside jokes with each other, but I've noticed recurring problems.
1. Only one person is always talking. This is a problem with many school plays too; often, a conversation consists of a lot of interruptions, or even someone making a remark in the middle of someone else's statement.
2. People are too audible. I realize this sounds ironic, but the thing is, not many things that typical people say are going to be said slowly and clearly. Games should try to establish which lines are more important for the player to hear (we're in an ammo cache, so WATCH your SHOTS), say those clearly, but leave other lines like "not that Biggs here is even listening..." in a slightly more slurred way.
3. It's always in the context of combat. Compare the level of interest of these lines...
"Ryan, you're ALWAYS the first to run out of ammo!" "It's called suppressing fire, moron!"

"Hamlet, act 2 scene 3, line 42." "Memorization like that means you know shakespeare to look smart, not just to like it." "Nah, I was just spouting random numbers anyway."

It's good to see another studio making a strong attempt at facial emotions, but I really feel like their methods are a bit too overpowered; motion capture is what you'd use to show a guy getting knocked to the ground. I sort of think that Half-Life 2's muscle system (having used it myself) is capable of doing just as much, much more easily. I could make something similar to that Heavy rain demo given a month of time, WITHOUT a team backing me up. (not to say that it's not still impressive!)

 

[ccesarano]

My problem is his "I'm not here to make video games, I'm here to make art" statement in the beginning. From reading that, his definition of art, then, is quite narrow-minded. Yes, it's highly skillfull to make people feel for other characters with facial emotions, but that's simply part of acting and drama. I know it isn't what he meant to say, but it just gives me the feeling that video games can only be art by copying other art forms.

I've always found game design itself an incredible form of art that you can't experience elsewhere, even if all you are making is a shooter. After all, one of the things Half-Life 1 & 2 has going for it is the level design. The beginning of HL2 takes you into this large, open city, and while you are actually following a linear path, they do a fantastic job of making you think you're being clever and finding hidden passages to outsmart the enemy. The first time I played it I wondered if there were any other options open, because it certainly seemed that way and I was merely taking the best one. To me, that's art. Same with the first Half-Life, where encounters with bosses were the entire level. Take the giant blind and green three headed monster. You never directly fought it, but instead traveled the whole level setting up its demise. Things like this I consider art, despite being in a shooter.

Obviously not all games are going to break the boundaries of art, but not all pieces of art do in the first place. Art is not just for the high brow, after all. Besides, if he's so reliant on things like facial expressions to bring out emotion and make a game recognized as art, then he'll never create as amazing a piece of art as, say, EarthBound, which got by without facial expressions.

 

[Arbre]

I still have hard times to adhere to his vision, a dedication to drive video games onto a path that is always presented as a trail of what other art forms or medias already are, or do.

His interview in EDGE, way more than one year ago, was already sweating with such ideas, and I already had issues with his claims.

I understand the point of having expressions that do generate emotions, applied to synthetized characters, helps to channel emotions.
A FMV featuring characters with shallow gestures or expressions would fail, as even the moves would not provide a message. They would not "talk" to us.

But do you really need ultra realism? Maybe it could help, but I don't think it's the objective #1.

I think the death of Aerith by the hands of Sephiroth, years ago, made that point fairly clear.

There is a problem with motion capture. I know it's pre-recorded data. Lots of it. Video games should not be about past information. It should be about present and future events.

Otherwise, what you're asking for is nothing more than an evolved FMV, and you want this [http://i2.photobucket.com/albums/y5/albumuser/misc/daedalus.jpg] mixed to that [http://i2.photobucket.com/albums/y5/albumuser/misc/janine-habeck-playboy-playmate-01-l.jpg], with some interactivity [http://i2.photobucket.com/albums/y5/albumuser/misc/ddr-screenshot.jpg].

Mocap only gets interesting when you make it dynamic [http://graphics.cs.ucr.edu/papers/zordan:2005RM.pdf], as it would answer a wide range of stimuli (pain, joy, concentration, leisure, effort, etc.), or sounds make the by characters or avatars.

The unease caused by artificial humans that are close to, but not entirely, realistic, is often called "the uncanny valley," a term coined by Japanese roboticist Masahiro Mori in 1970. According to Mori's theory, a replica of a human will elicit more empathy the more realistic it is, up until a point where the simulacrum becomes so lifelike that small faults pop out and make it seem uncanny. The "valley" is a dip in a graph that describes the level of familiarity one feels with the simulacrum.

Makes me think that the closer you'd get to the perfect reproduction of the human body and soul, the higher the chances would be you'd get crazy.

But I think the uncanny valley applies to the AI, not the enveloppe.

Does she [http://i2.photobucket.com/albums/y5/albumuser/misc/japanese-sex-robot-790079.jpg] looks uncanny?

What would rebut more would be how bad people could react to machines. It's not that I'm particularily a pro-android guy, but that scene in Animatrix were a bunch of mobs smacked down a robot for just being a robot in human appearance, that was gross. Or grotesque.

Really, I'm not convinced that the uncanny valley applies to visuals. No matter how realistic looking your virtual human will be, it will be nothing more than a fake.
I don't think that if that gal [http://i2.photobucket.com/albums/y5/albumuser/misc/final_fantasy_x2.jpg] started smiling or crying in real time, I'd be moved, not if the actions that led to those emulated emotions were meaningless. The context would play much more than the content.

Now, put a near human intelligence in that bag of pixels, and then I say you'd get something uncanny. In fact, you may not even need a realistic looking body for that. The mere idea of talking to an AI that could reply to you and keep the discussion going on, would probably unsettle you a thousand times more than a mass of polys that can "cry".

"If we can make simple scenes from daily life interesting to play, like two people just talking, then we have a whole new world in front of us," he says. "Then we can do anything."

Possibly, but I don't think he's putting the emphasis on the right objective.
There's an important part that makes the difference between an interesting discussion and a boring one: the sentient and intellectual bond that exists between individuals.
This has hardly anything to do with what they look like. What he wants is a humanoid AI, not an empty shell that looks human.

 

[brkl]

No amount of motion capture would have saved Indigo Prophecy's travesty of a storyline. If David Cage wants to make art, that's where he should concentrate.

 

[Arbre]

He should concentrate on getting that damn sequel to Nomad Soul out.

He's got a huge studio, and doesn't release that many titles. QD is 10 years old, and only has two titles in its catalogue. There's been more than 5 years between both of them.
Besides, they're all flawed to several extents.

I'm afraid he still has the mentality that led many French studios to their downfall, "art blah blah", with the difference that for some reason, he's got a lot of bucks coming from god knows where, maybe his mocap studio he rents to various other studios, which helps him keep his pompous pretense without loosing face.

Fact is, no matter how acclaimed his two games were - or not - they're hardly cited as references.

 

[Lance Icarus]

Facial animation can be very expressive when it comes to conveying a character's emotions and personality, but concentrating so hard on one minuet aspect of video game presentation means that some aspects of the game are pretty much doomed. So your female lead seems to give off a warm feeling with her smile like a mother welcoming in her child from his snowy escapade, too bad her dialog sounds like it was written by a 13 year old Valley Girl.

Before you start getting high and mighty about how your characters express themselves through the way they look, you have to make sure they have something to express first. Otherwise, it's the speech that seems jilted and unrealistic instead of their face.