reptile said:
On the other hand: GOW has the _worst_ nextgen particle stuff i've seen. It's amazing that with tools like that, they cant manage to hire someone to do at least decent effects.
Yeah, whereas MGS has very nice atmospheric effects, the smoke/dust blowing in the wind adds a lot of punch to it.
Okay, now seriously, I'll try to turn this discussion to an objective comparision based purely on the technical stuff and trying to keep the subjective part out of it.
Characters. Technology wise, both engines are using simple polygon based models (no HOS to smooth out the curves, definitely no displacement mapping), and normal mapping to add further detail. Both use soft self-shadows, probably implemented through shadow buffers, on the characters, although MGS seems to have sharper shadows.
MGS spends a lot of polygons on Snake's character, which is understandable considering the amount of close-up shots in the cinematic. The actual ingame scenes will probably use a lower detailed LOD simply because you wouldn't see the details anyway.
60.000 vertices for the whole model sound pretty much right to me - you can see a few poly edges if you look hard, and some of the detail is only present in the textures. There are a few dozen individual hair strands which however fail to blend with the rest of the hairdo that's a solid piece of geometry.
Soldiers are also quite detailed with straps and pockets individually modeled, and so is the little robot with some actual hidraulic pistons modeled into it.
Texture wise, it's not that bright though. The amount of detail is a bit lacking, there are large homogeneus areas in Snake's armour and his skin is quite monochrome as well. No sign of dirt and damage on the suit, and the same goes for Otacon and the soldiers as well.
Shader wise, there is some skin shading going on, but it looks more like plastic with some glancing shine thrown in that is not affected by the shadows (that's why Snake's face has that outline someone mentioned before). The suit looks quite dull, it's actually pretty similar to Doom3/Quake4's marine armour.
Character animation seems to be standard bones and skinning, with probably some extra bones thrown in to correct the bad looking deformations (particularly on the shoulders). Working with the PS2 for so long, Kojima's character riggers should be quite experienced in this. Sorry, no muscle simulation or anything here, those technologies are far too expensive to compute even for a model with half of Snake's resolution (like the Cave Troll in the LOTR movies that was ~30.000 NURBS control vertices). The simple secondary dynamics on the head scarf were already done on this gen of consoles as well, though they add some nice movement to otherwise static shots.
GOW has a lower polygon count in the characters, although it's considerably higher than the UT2007 ones. It's probably around or above 10.000 vertices, and a lot of the detail is pushed into the normal maps extracted from some highres Zbrush digital sculpt. They probably don't plan to do super close-ups, although I don't know how many characters they can display at once (a larger number of enemies would be a good reason for this compromise, yet we haven't seen more than 10 at once).
I would actually think that one of the main difference in the look of the characters is that Epic uses Zbrush to add lots of detail to the highres geometry for the normal maps, whereas Kojima's team is just building highres models in a 3D animation package. Now, in Zbrush you can tesselate your geometry to 2-8 million polygons and sculpt each of them individually using brush-based tools; whereas in an animation package, you build a subdiv model that gets tesselated and smoothed, but no further detail is added. This explains the smoothness of the MGS character faces, and the look of Snake's armour. Just look at Otacon's face in the highres shots to see what I mean.
Textures are a lot more detailed, wear and tear and dirt are very evident (look how all the edges of the rifle are more shiny, to show that it's been used and worn down). Generally speaking, Epic's style is hyper-realistic, meaning that the amount of detail mimics real life but everything is exaggerated. The same goes for shaders: specularity is very strong, the fake subsurface scattering is almost overillluminating the skin on the characters' faces. I'd say that it looks like a Michael Bay movie, Armageddon for example. MGS on the other hand is far more stylized and has a painted look. It also effects the way the scenes are lighted - MGS is monochrome again, with low contrast and at sometimes quite flat enviroments. GOW on the other hand is exaggerated again, strong key and fill lights with high contrast in lit/shaded areas, amplified by cool/warm color variance (ie. the good old orange/blueish pair). Tehcnically, it's shadow buffers and point lights, although MGS may be using some trickery on the secondary lights. But the result does not look like HL2's bumped radiosity and ambient cube lights, though, more like Outcast's software rendered bumped characters. GOW has a small advantage in the precalculated shadows for the enviroment, but there's not such a big room for variance in lights anyway.
Vehicles are comparable to characters - the low res textures on the tanks have been mentioned already.
Enviroments are far more detailed and complex in GOW. The amount of geometry and specific normal maps is astounding, however, everything is static. MGS has all the smoke, dust, and debris that make it feel a lot more alive. Epic should look into atmospherics as well, one of the cool things in the original Unreal was fake volumetric fog after all.
Note that the atmospherics in MGS are 'fake' as well, that is, no true volumetrics but textured particles mostly. I hope that I'll be able to link to a nice example image sometime next week to show what I mean... anyway, textured polygons are a prefectly OK way to do these effects - the fire of the Balrog and much of the water flooding Isengard in LOTR movies were utilizing the same techniques (although they were able to use animated textures - real footage - because there wasn't any memory constraint). Spending R&D on realtime volumetrics is not justified IMHO.
DOF and motion blur are both image processing effects, currently it seems that MGS has DOF, GOW has DOF and MB. Not much to be impressed with, offline rendering has used both effects for many years now. MGS also seems to have a constant color correction process going on that creates the green-yellow tint. This also adds a lot to the cinematic-ness of the trailer.
So, to sum it up:
- similar technology in displaying geometry: normal mapped polygons
- similar technology for self-shadowing: shadow buffers
- higher quality precalculated shadows for level geometry in GOW
- MGS has more polygonal detail, GOW has more normal mapped detail (particularly in enviroments)
- smooth skinned skeletial animation for character bodies and faces (ie. more than 2 bone influences per vertex)
- background streaming of data in GOW is to be tested in practice but sounds pretty interesting
- better quality in textures in GOW
- better use of particle/atmospheric effects in MGS
- more dynamic geometry in MGS (flying debris, blowing smoke - no physical simulation required for these, simple particles)
- better use of image filtering effects in MGS
- biggest difference is art direction: GOW is hyper-real, MGS is somewhat painted but with many more 'realistic' elements implemented (particularly the atmospherics)
Practically all of the above are current generation technologies, but they haven't been utilized at the same time and to this extent in games that have been released.
I'd expect both to advance a bit until their release, MGS in particular because they've been probably working with a tight schedule for this trailer. But it will also depend on what the designers want - Kojima might be content with the current texture quality, for example, because he's more interested in the cinematic effects like atmospherics, image processing, etc. GOW would benefit a lot from adding such effects, too. It seems that Kojima has the art director with more experience or knowledge in offline CG, because they're implementing many commonly used and relatively simple offline techniques that can have a surprisingly big 'wow' effect.
And in the end, people should stop arguing about which is better, because B3D is not the forum for this - we should concentrate on the technical side of things, how they're doing it, what are they doing after all, and so on. Try to be friends like
these two guys here...
