AMD Cinema 2.0 tech demo: real-time photo-realistic human models

by Andrew Yoon { Aug 12th 2008 at 12:01AM }

click to view gallery

At first, this image may seem rather boring. Who's this woman and why is she on the front page of Joystiq? This image is rather uneventful until two facts are revealed: a) she's a CG creation and b) she's being rendered in real time with new AMD technology. AMD's new effort, "Cinema 2.0," promises to bring photorealistic graphics that blend the visual fidelity found only in the most recent of Hollywood movies to the interactive space. AMD is imagining a future where consumers won't just play movies, they'll play in them.

While practical applications of "Cinema 2.0" weren't showcased at AMD's recent event in New York, the company executives were keen on noting how the tech can be implemented. Movie producers will be able to manipulate digital actors that are more realistic than ever -- in a rendering environment that happens in real-time. Gamers will be able to find character models and environments that are truly lifelike.

Gallery: AMD Cinema 2.0

The demonstration we saw in New York did prove that this technology is very real. A very human model, one that would put the efforts of Square Enix to shame, had all sorts of dynamic lighting applied to it. But how is it that AMD has managed to create such a realistic human character model, while Hollywood still struggles to create true human actors with a gigantic rendering pipeline?

Obviously, new hardware is part of the solution. More processing power has always meant more realistic visuals. However, there's another trick: a globe-shaped room that feature rapid high-definition photography from all angles. These cameras can capture surface detail to a level that surpasses human touch. These cameras operate so quickly that to the naked eye, they appear as solid lights, when in fact they are flashing on and off at fractions of a second. The information captured by these cameras can then be interpreted into 3D data. Here's the same woman we featured as these cameras capture her:

Once the digital actor is captured, directors (and game players) will be able to manipulate them how they see fit. Some truly science fiction applications were mentioned during the presentation, like the ability to recreate Star Trek's Holodeck. We're not holding our breath for that, but this stunning technology does appear to be a revolutionary leap forward in rendering technology.

While AMD will tout that this technology is available "now" for everyone, that is far from the realistic truth. As with all advancements in 3D technology, the availability of new graphics hardware doesn't necessarily mean they'll be utilized by developers. For example, how many games push PC graphics cards to the levels of Crysis ... other than another Crysis?

Developers from Splash Damage, Crytek, Rebellion, Remedy (yes, they're still making Alan Wake) and even Blizzard were highlighted in the presentation as partners of AMD. While it's clear that this impressive technology will be used, the real question is when will gamers see real playable products on store shelves? That may be years off ... and who knows if competitors will be able to produce equally impressive visuals on an even larger scale by then? Until then, enjoy the eye candy.

Hundreds of high-definition rapid-firing cameras help create the impressive 3D models used in the AMD demo.
Images Provided by Paul Debevec and the Institute of Creative Technologies Graphics Research Lab

Tags: amd, ati, breakingnews, cinema-2.0

Related From Joystiq

Ubisoft envisions a future of 3-D movies, games
9 days ago
PSP 3D viewer release appears slightly more distant
15 days ago
CES 2009: Sony shows off 3D display tech with PS3
336 days ago

Reader Comments (Page 1 of 2)

Ciatrick Leodis McKelvin @ Aug 12th 2008 12:04AM

What uncanny valley?

fwacce @ Aug 12th 2008 12:12AM

Please don't bring that phrase up. It has become the most overused and completely irrelevant term on the internet as of late. Every time there's any posting about CG people, the conversation has to be about UV to try and appear smart. It's just a theory that may or may not be useful, but so far unproven as to whether or not it even applies.

danielhardy001 @ Aug 12th 2008 12:30AM

Eh? Uncanny Valley is not irrelevant, it is a very real phenomenon.
Just go see Polar Express, that CG xmas movie from a couple of years back. The people look very realistic, but have creepy dead eyes which make them quite freakish.

Also, see some of the real-world robots that are made up to look cosmetically like humans. If they are still, they look human, but then they move, it is slightly wrong, and they look extremely creepy.

Ciatrick Leodis McKelvin @ Aug 12th 2008 12:39AM

@fwacce

This looks like it IS crossing the uncanny valley. We don't even see that realistic CG in movies. Don't be such a twat about it.

Exo @ Aug 12th 2008 1:06AM

@ fwacce

quite the contrary, it looks like you are trying to appear smart by refuting the statement and acting as if you know more, enough so to say that this topic is overused.

just shut up if you don't want to be a part of discussion k?

Paul (PSN: heypaul) @ Aug 12th 2008 1:54AM

Quit being such dicks to fwacce, guys. There's a reason the uncanny valley is just a hypothesis. The things that we consider outside the valley today (for instance, over-exagerated Pixar CG characters) are the very things that Mori would've considered smack dab in the center of the valley when he proposed the hypothesis in the 70's. If you choose to argue the existence of the valley at all, you'd have to at least concede that the perception of where this valley exists shifts with technological advancements in portraying artificial humans. You may think Buzz Lightyear is too unrealistic a representation to fall in the uncanny valley, but I guarantee you, it'd creep a Yanomamo tribesman the hell out.

SkreeHunter @ Aug 12th 2008 2:04AM

Hey nerds. Shut up.

BananaBoat @ Aug 12th 2008 2:19AM

You can beat the uncanny valley in still images. Video is a completely different story. Personally though, I don't think they've even beat it with a still photo this time around (the girl in the chair looks very..well...uncanny) (ultimate lulz awarded if that is actually just a photo of a girl, and this is all a trick, and she really is just that uncanny IRL)

Anyway, none of that really matters. What matters is that this technology is completely pointless. Why do I want to interact with the movie I am watching? It's cool in a game like MGS4 where you get to switch viewing angles and what not...but for a film, I'd much rather have the director pick the BEST angle, without me doing anything, so that I won't miss anything and so that I don't have to screw with it. I realize the gaming implications of this, but it's completely impractical from the gameplay perspective. This seems no more interesting than the previously mentioned MGS4 interactive cut scenes, and it feels like a gigantic waste of time on AMD's part.

Maybe I'm not seeing the big picture, but other than for cut scenes, I don't see this getting any use in actual gameplay. If moving the camera while your character is on rails is of interest to you...then...well that tech already exists. Pixar could have announced that you can now pan and tilt the camera in it's next movie, and it would have been about as equally exciting of an announcement.

Alcevious @ Aug 12th 2008 5:14AM

Great, now every nerd who's used the phrase "uncanny valley" is going apeshit on fwacce.

NoHitHair @ Aug 12th 2008 6:39AM

The uncanny valley argument is largely irrelevant for any practical reason. Beyond that, the evidence supporting it is contradictory and circumstantial at best.

BananaBoat @ Aug 12th 2008 6:48AM

The thing about the uncanny valley is...you either see it, or you don't. Trying to explain it to someone that doesn't notice it, is like trying to explain the color blue to a blind person. (Sorry if anyone is uhh...having a text-to-speech thing read this to them) It's equally as idiotic for someone that doesn't notice it, to try to tell someone that does, that it doesn't. Who are they going to believe? Their eyes, or you?

Personally, I can't watch things such as the CGI movie Beowulf or the CGI movie Polar Express. The character models freak me the hell out with their massive valleys.

Burnt Meatloaf @ Aug 12th 2008 7:09AM

I'm not sure if the uncanny valley even applies, here. If everything is put together with camera shots, then what you're really seeing is a reconstruction of a real person based on photography. The uncanny valley shows up when people try to build their own models and animate them with more artificial means, such as point sample motion capture or hand animation.

This is more like putting a 3D photograph into a scene, airbrushing it, and adding custom lighting. Impressive, and it requires real horsepower to do it in realtime, but I'd rather see something done from scratch. That's harder to do, and more representative of real, practical products.

Wait... did I just say "practical" in a thread about GPUs with teraflops of performance? Never mind.

Ghen @ Aug 12th 2008 7:24AM

The uncanny valley doesn't exist. Just people not finishing the work they started.

"Oh its good enough, send it to the developer and lets have lunch"

Freelancepimp @ Aug 12th 2008 1:22PM

These comments are an example of what the guys who know everything do when no one in the real world wants to be around them anymore. They just come here and piss each other off.

Korova Pamplona @ Aug 12th 2008 8:45PM

Wow, that is one well-fed troll.

GreyFox @ Aug 17th 2008 2:47PM

Hey!

I liked Polar Express!

mose @ Aug 14th 2008 3:45PM

i'd hit it.

blehh @ Aug 12th 2008 12:06AM

amazing

FOXHOUND @ Aug 12th 2008 7:25AM

...This just inches us that much closer to S1M0NE, eh?

Truth be told, I'd prefer digital actors & actresses over the actual ones. It would put all those stupid TMZ-ish shows to the wayside, because the creators could just "delete" any character they don't need. ;]

mapi @ Aug 12th 2008 12:09AM

The real intended application of this technology?
Porn.

Ciatrick Leodis McKelvin @ Aug 12th 2008 12:11AM

hell yeah son

Marilyn Monroe is gonna be gettin the business by John Wayne

KeenCommander @ Aug 12th 2008 1:12AM

And there goes your child pornography laws...doh!

klitorisaurus @ Aug 12th 2008 1:27AM

I can't wait to see a Whoopi Goldberg scene!!

Tehman @ Aug 12th 2008 12:16AM

So this was what Valve was waiting for to make their Gordon Freeman model as realistic as possible....

Arkon @ Aug 12th 2008 12:18AM

Screw pictures, see this stuff in motion at youtube under "cinema 2.0". I hope this technology reaches the next generation of consoles although I still have doubts. Seriously though, fracking awesome!

Solid Chief @ Aug 12th 2008 12:17AM

Odd that it makes her look older...

Huey2k2 @ Aug 12th 2008 12:17AM

Holy shitcakes... I didn't realise that wasn't a real person when I first looked at it.

Halwende @ Aug 12th 2008 12:22AM

The Witcher 2: Enhanced Edition confirmed for 2010 with Cinema 2.0 technology! :P

zuburi @ Aug 12th 2008 12:44AM

Continuity across different posts? You win, sir!

Sesur @ Aug 12th 2008 12:24AM

Looks phenomenal for CG; looks crappy as a image.

Grey Acumen @ Aug 12th 2008 12:32AM

that's because in order to cover up the "uncanny valley" effect, they dropped the image quality down and put it through a few filters to make it look like an amature webcam or photo. This way you don't have the detail to recognize the ways that she looks off.
If you look at the better quality pics, you can see that it's still inside the uncanny valley, though I will admit that they're on the far cusp of it.

fred @ Aug 12th 2008 4:08AM

Filters my butt, do you know how much power it takes to render? Realistic light algorithi-thingies and jiggawatts and such take forever.

You have to tighten up the graphics on all the levels up to 11.

I should know what I'm talking about: I'm on the internet.

porath @ Aug 12th 2008 12:28AM

are you guys sure the image in the main body is 3d rendered? the other examples in the gallery are quite obviously of lower quality than the first supposed rendering...

porath @ Aug 12th 2008 12:32AM

nevermind, in the youtube video the guy says it's actually rendered in realtime, but i wonder why that other character looks so much worse in comparison

Grey Acumen @ Aug 12th 2008 12:36AM

See my comment just above yours where I was replying to sesur. The first image they show is actually a lower quality pic.
Essentially, they are just simply NOT showing us the detail, so that our own mind will work to fill in those details. We'll fill things in so it actually looks right. The other images have more detail, so our minds don't attempt to fill anything in and it can see which details actually make it look less than real.

the_insider @ Aug 12th 2008 3:42AM

The difference is that in the first shot, all you are seeing is a hi-res 3d captured head and nothing else. It's canned, captured, data, with nothing but a 2d environment map background. Obviously, they are able to push the levek of detail a lot more and the fact that it is pure captured data makes it a lot easier to get the 'realistic' look right.

In the second sequence from the new Ruby demo you are actually seeing a whole 3d environment. It's still nowhere near as expansive as a real game but at least there is a full uni-directional set (if the camera would swing around there would be nothing, as witnessed in one of the demos). There are also prop cars, two moving characters, more shaders, etc. In short, there is a lot more going on. Hence the quality bar drops slightly, and on top of that Ruby has a really cartoony design so she sticks out like a sore thumb from the otherwise more photorealistic environment.

Personally, I am actually a lot more impressed with the second set of images (and the demo they are from) as they actually provide a real glimpse of what next generation games are going to look like. The environment, detail, lighting, shaders on the cars, and post-fx certainly look a generation ahead of what we are seeing now and by the time we actually get games that really take advantage of this gfx cards will have advanced yet another generation or two. To imagine a fully fledged game with AI, more animated characters, vehicles, etc. on screen at this sort of fidelity (and with a more unfified art direction) is rather exciting.

As a side note, that capture technology is nothing new really. EA have been using a similar approach, based on Paul Debevec's work, for capturing facial animation for games like Tiger Woods and Need for Speed, and probably others, for a while now. The final in-game quality however, is not as high as in this highly streamlined tech-demo but that is more a limit of the current generation hardware and the fact that it is running in a fully fledged game environment.

Peace.

DivinoAG @ Aug 12th 2008 10:10AM

I'll just start by saying that I have seen this technology video quite a few years ago, after it was released on a Siggraph; that's not brand new developments.

The point of this particular tech is to allow realtime *re-light* of a captured video, not of 3D models. So unless you're doing a live-action game, the technology demonstrated here isn't going to help much. And since it's live-action, no uncanny valley concepts apply here.

Despite that, this was one of the visual effects breakthroughs that allowed the development of techniques used for the Matrix Reloaded, particularly on the Burly Brawn sequence (now you know how many years ago I've seen it), which was latter much improved to use with Doc Octopus for Spiderman 2.

What those techniques allowed directors to do was film the actors doing the face performance that would appear during action sequences, and that performance would then be seamlessly (as possible) integrated with the body 3D models.

In conclusion, the only thing this article offers new is that AMD is working on a hardware solution to make all this run in real time. But there are still many years until you see that showing up on a new Call of Duty.