I actually know very little about what they're up to with next gen. Im destined to work with the PS2 til I die EA Canada seem to be in charge of all the cool next gen stuff so Whargoul might know more about this.
Impressive. I'd put money on that being a cyberscanned head.
It's interesting that it doesn't appear to be normal mapped. I dont see any polygonal edges. So it seems they're thinking about skipping nmaps and going straight to high res geo? I guess for a boxing game there'd be no problem storing a couple models this high res in memory on next gen hardware. Or perhaps theres some realtime Subdivision happening.
Care to elaborate Cody?! Or is this stuff not meant to be public domain yet?
The texturing is really impressive. I have doubts about that being a cyberscan. Sure the geo could be... but the textures aren't anything anyone would want to work with. Plus the texture itself doesn't seem to have that photo-slapped look. It looks like the lighting is doing quite a bit of work.... coupled with multipass textures along with it.
The geo could be high res... but I doubt it's subdivided. It doesn't have to be ultra smooth when you consider how low res TV resolutions (and those images) are.
Even with TV resolutions and not great image quality allowed for, the density of geometry in say the ear, is very apparent and way beyond anything currently around.
The diffuse texture is photographic in source. Of that Im positive. I think the most impressive thing happening here is lighting. Whargoul, where are you?! Questions need answering!
It looks pretty similar to the Matrix-style scans where they grab the displacement and spec maps at the same time as the diffuse.
Looks beautiful, but it's not a game, it's a floating head that doesn't appear to animate. I don't doubt it will look this good in the fighting game, but when you only have to draw 2 characters it's not hard to spend alot of resources making them look good. That grain of salt may have been obvious, but I thought it was worth saying.
Next up, high poly gaming. I bet displacement maps are the key component here. A lot of great results with that now that Zbrush is here. This guy uses it very well.
Not sure what I can really say here, but who cares? most of it is guessing anyways. At one time I did have these assets on my machine and was checking them out (Rendermonkey project), but this is newer and better than that stuff.
Yes, that is a real asset, not a photoshopped head. The base geo is pretty damn high, but it is normal mapped (see the pores and shit on the face?). I think the base high rez head was created from a cyber scan of the guy, and I seem to recall something about them doing a lifecast of his head and scanning that as well? There's this new guy that worked on data capture for the Matrix films, and it looks like his input is helping out quite a bit. Base texture is probably from photos/cyberscan and hand cleaned up. Spec maps probably hand painted. Normal maps generated from high-rez low-rez difference.
As for the lighting, it's most likely per-pixel spherical harmonics. The skin shader looks like it uses texture space diffusion (a sub-surface scattering with "blood transfer" fake), and it doesn't look like they're using the translucency fake (no light shining through the ears - someone else had a really nice demo of that I think). Blooming may be on, hard to tell. There may be an ambient occlusion map used in the lighting as well, it's hard to tell in those shots. It would likely be generated off the mesh as well. Looks like there's no shadowing in there - too bad, it could make it look really nice.
Background was probably an HDR probe capture of the enviro they shot the guy in, and the SH map was extracted from it.
Keep in mind this is all just educated guesses of course, I don't work on this project.
Assets dont affect gameplay. Just instead of doing them in-house you buy them and then pretty much drag and drop.
Actually then all realistic games look just about the same and gameplay/story will become more important than now.
I still think UE3 looks better. Its just nice and blurry and low res. If I recall, anything on a TV screen that is videotaped will appear to be ungodly good but once to get it on a high res monitor is just a mess o' ugly.
Hmmm...good eye candy. But you never know if that single floating head is using up all the resources of the machine, leaving no room for AI, backgrounds orother game logic
As far as the big words go: I got most of them, except 'ambient occlusion map'. Now I know someone will have the urge to point me to www.justfuckinggoogleit.com, but in the interest of all lazy bastards here, can someone explain that one?
Ambient occlusion is sort of like a global illumination/radiosity lighting calculation, only its based on light emanating from everywhere at equal intensity as I recall, as opposed to a directional lightsource.
Something like that, although people are getting to the point that they are not baking it into a texture once, but generating the texture every few frames, thereby having 'realtime GI'. A while back there was someone here who was working somewhere, where they were doing that. It's in the thread about skin color, not sure if that is still reachable.
Ambient occlusion is used in conjunction with a global ambient light source such as a cubemap or sherical harmonic map. It defines what part of the geometry receives ambient light, or general global light from the environment. Forms self occlude, so you'd get less ambient light in areas like behind the ears and on the forehead.
They're definitly using floating point precision for their lighting, similar to what U3 is doing. We'll all be moving to this format eventually, it's way to powerful to ignore it for long. True high dynamic range imagery is the future!
Replies
It's interesting that it doesn't appear to be normal mapped. I dont see any polygonal edges. So it seems they're thinking about skipping nmaps and going straight to high res geo? I guess for a boxing game there'd be no problem storing a couple models this high res in memory on next gen hardware. Or perhaps theres some realtime Subdivision happening.
Care to elaborate Cody?! Or is this stuff not meant to be public domain yet?
The geo could be high res... but I doubt it's subdivided. It doesn't have to be ultra smooth when you consider how low res TV resolutions (and those images) are.
Still incredibly impressive, though.
The diffuse texture is photographic in source. Of that Im positive. I think the most impressive thing happening here is lighting. Whargoul, where are you?! Questions need answering!
Looks beautiful, but it's not a game, it's a floating head that doesn't appear to animate. I don't doubt it will look this good in the fighting game, but when you only have to draw 2 characters it's not hard to spend alot of resources making them look good. That grain of salt may have been obvious, but I thought it was worth saying.
http://www.taron.de/Neckling.htm
Yes, that is a real asset, not a photoshopped head. The base geo is pretty damn high, but it is normal mapped (see the pores and shit on the face?). I think the base high rez head was created from a cyber scan of the guy, and I seem to recall something about them doing a lifecast of his head and scanning that as well? There's this new guy that worked on data capture for the Matrix films, and it looks like his input is helping out quite a bit. Base texture is probably from photos/cyberscan and hand cleaned up. Spec maps probably hand painted. Normal maps generated from high-rez low-rez difference.
As for the lighting, it's most likely per-pixel spherical harmonics. The skin shader looks like it uses texture space diffusion (a sub-surface scattering with "blood transfer" fake), and it doesn't look like they're using the translucency fake (no light shining through the ears - someone else had a really nice demo of that I think). Blooming may be on, hard to tell. There may be an ambient occlusion map used in the lighting as well, it's hard to tell in those shots. It would likely be generated off the mesh as well. Looks like there's no shadowing in there - too bad, it could make it look really nice.
Background was probably an HDR probe capture of the enviro they shot the guy in, and the SH map was extracted from it.
Keep in mind this is all just educated guesses of course, I don't work on this project.
Now that makes me wonder how long the next gen games will take to develop. Doom 3 took what? 5 years?
[/ QUOTE ]
Probably really short err *snicker* "games" then..
Now that makes me wonder how long the next gen games will take to develop. Doom 3 took what? 5 years?
[/ QUOTE ]
Asset libraries are the way to go. Movies get away useing same actors so why shouldnt games do it?
Actually then all realistic games look just about the same and gameplay/story will become more important than now.
As far as the big words go: I got most of them, except 'ambient occlusion map'. Now I know someone will have the urge to point me to www.justfuckinggoogleit.com, but in the interest of all lazy bastards here, can someone explain that one?
They're definitly using floating point precision for their lighting, similar to what U3 is doing. We'll all be moving to this format eventually, it's way to powerful to ignore it for long. True high dynamic range imagery is the future!
It's a very cool demonstration of tech.