Now that's definitely going to need some sort of backing up. What is that supposed to mean? The geometry on the elite is completely alien with the curvature and varying degrees of angles. It's a wonder that you don't see any UV stretching or warping on any portion of the texture. Curvature is even tougher to handle with texture compression artifacting. The picture of the elite there couldn't be less complex. Even the weapon it's holding shows a high degree of resolution and complexity versus a more angular weapon.
Range of graphics effects in console games *spawn
- Thread starter Cyan
- Start date
Now that's definitely going to need some sort of backing up. What is that supposed to mean? The geometry on the elite is completely alien with the curvature and varying degrees of angles. It's a wonder that you don't see any UV stretching or warping on any portion of the texture. Curvature is even tougher to handle with texture compression artifacting. The picture of the elite there couldn't be less complex. Even the weapon it's holding shows a high degree of resolution and complexity versus a more angular weapon.
I mean the surface of the outfit (shadow, bend etc etc) implied more complex details in other graphic department.
What does simple mean? Simple designs? The technical complexities of normal mapping a seruies of regular squares or complex veins and wrinkles are exactly the same. The only differentiating factor regards quality is normal map resolution, compression, and the filtering settings. Comparing the impact of normal maps in one game versus another, it's pretty much just artistic appreciation that people end up comparing. Unless you can point to the normal maps of Reach being lower resolution or somesuch, you have no argument other than, "I don't like the look of them." ANd that seems to be the basis for much of the past page of talk - very little actual metrics and supporting, evaluative references, and a lot of "this looks nice and this doesn't." There's nothing wrong with having preferences and liking one game's look over another, but it's a folly to talk about technical differences where there are none.
Eh, edited myself out.
Anyway, Reach obviously has very high res textures, but the detail work is subtle and relatively low contrast when compared to the sharp, almost black and white look on KZ.
Which one a person likes is a matter of taste and we can't argue about that. But questioning the technical achievement is stupid; Bungie has created characters with a LOT of detail and there can be no denial of that.
Anyway, Reach obviously has very high res textures, but the detail work is subtle and relatively low contrast when compared to the sharp, almost black and white look on KZ.
Which one a person likes is a matter of taste and we can't argue about that. But questioning the technical achievement is stupid; Bungie has created characters with a LOT of detail and there can be no denial of that.
I guess use different complexities on the outfit implies different technic, no more. I mean, use wrinkles or complex veins maybe implies more self shadowing & 'dynamic' shaders to simulate the physic effect? However with simplistic I'm not try to disclaim the graphic department of Halo reach... to the opposite I read a lot of criticism of kz3 engine. I'm not try to said it isn't legitimate but we have to know how exactly work & even how smart could be someone, without concrete pdf, the argument remain matter of opinion anyway.
Last edited by a moderator:
Lucid_Dreamer
Veteran
That's right, you did say you had a triple GPU setup! That's a beast of a setup! I agree with you, though. The look is very similar.Killzone 3 isn't that far off Crysis in terms of clothing detail...
1920x1080, DX9 Very High Config with 8xMSAA+8xTRSAA - Nice frame rate to boot
Image shack has killed the detail but Killzone 3 uniforms are very close...
Are you saying that's what KZ3 is doing to achieve a look that's similar to Crysis (PC) textures? If so, why aren't other console games showing this look? If you say there are console games showing a similar look like Crysis (PC) and KZ3 textures, please post pics.
Question: Does Halo: Reach stream their textures throughout the entire game or is it loaded into memory for each level?
Last edited by a moderator:
Would I be wrong in saying the texturing techniques for Killzone 3 and Halo Reach would essentially be the same? The differences would be aggressiveness of LOD transitions, resolution of base/normal/detail and the most apparent differences would be the actual artwork?
I'm just curious to know how different one game might be to another, in terms of how they do that kind of thing.
I'm just curious to know how different one game might be to another, in terms of how they do that kind of thing.
You dont understand what I am saying, I am not refering to the texture resolution. I am refering to the amount of textures and normal mapping on the particular character. And I am not referring to the quality either. I am saying that the particular character doesnt have much to showcase in terms of normal mapping due to the art. And yes comparatively it may have higher resolution textures compared to Killzone 3 but it doesnt have much anyways.
The image you posted doesnt show well what your saying and I did say that this is due to the art chosen on the character.
I am suggesting that you should pick another better image that shows what you are trying to say, specifically about the normal map detail. There is more going on with specular lights than normal maps on the character.
L. Scofield
Veteran
Wrong. Where in those shots does the cloth texture look like a pixelated mess (like KZ3's)? Crysis is using a high resolution base and detail mapping.
If you think of the texture as a sheet of paper that you had to wrap onto a figure, it probably becomes a lot clearer as to what we mean by complexity. The model itself will have points where it stretches more in one place than others and you'll see the effects of that in-game most easily at changes in polygon angles. Laa-Yosh could probably explain it better.
With the varying texture formats available, the compression artifacting will differ in obviety. i.e. Normals, Specular, albedo.Also note there are texture formats available on Xenos that aren't available on RSX.
Yes. Generic patterns using UV tiling (repeating) and maybe some masks to avoid putting them on everything.
Art direction mostly; other art guys might not prefer to paste tiled textile patterns on everything.
KZ3 is probably only resorting to it because they were unable to increase the texture budget per character from KZ2, but there's a lot of close-ups in cinematics and because of the melée killing moves and the lack of resolution would be disturbing. Especially on NPCs who have relatively large face textures in the cinematics.
Also, I'm not sure how costly it is to render multiple texture passes for characters. I'd guess it's only the normal maps that have a detail texture on top and color/spec is unaffected but it's still more expensive computationally than using a higher res texture.
Already posted a Reach shot; I think you can google Gears 2 and Uncharted 2 yourself...
It probably does, they also have a very good LOD system and the impostor system as well to conserve memory.
Yes you would be. Reach uses a single texture layer for color, normal, specular channels.
KZ3 uses a base texture layer with a relatively low resolution, then it uses another layer that has a repeating pattern which is overlayed on top of the base texture for at least the normal channel (because it affects lighting). It's maye also in the color map but I'm not sure.
Reach seems to have a 2K texture or so per character, maybe two. The detail is very high, you can see it on how sharp the textured lines are on the Elite's armor, there's barely any pixelation.
KZ3 seems to have a 1K texture and the detail texture, plus for humans a separate face texture; I'd guess another 1K for the main player character or so. I can check the D'Artiste book at home on KZ2, I think the sniper in that game had two rectangular UV layouts, but I'm not sure on this one.
This is how it's done:
The distortions can be reduced, today we have very high quality unwrapping algorithms. Making 95% use of texture space by manually arranging the pieces is still better left to an artist; you also need to leave some padding so that colors don't bleed over the borders with MIP mapping.
Texture detail in my definition is the size of a texel in, say, centimeters.
You can "cheat" by mirroring parts, like use the same texture for both legs/arms so that the overall texel density can be increased. You can also use multiple textures per character, use separate maps for heads, and so on.
Adding a detail map on top isn't equal to using a higher res main texture. It's also better than the lowres version, so it's somewhere in the middle. I personally don't like it because it's very evident to my eyes but that's a question of taste.
Also, KZ probably reserves significant texture memory for the face textures of the hero humans, and so has less memory left for the individual bodies and enemy types. The G-buffer also takes up a lot of memory, as well as the particle buffers and so on. So it probably has less texture RAM compared to Reach at the start, and then chooses to utilize it differently.
I can't respond to that, there's no such thing as "amount of normal mapping" in my book...
I've explained more of my points in the previous posts.
Yep but if you check the top left I'm only using 2 of them
Crysis is not the hardware hog it once was...
Laa-Yosh I'm sorry dude but I've not seen any other console game apart from KZ3 that can get close to Crysis models..
Thanks for the responses, Al and Laa-Yosh. It's nice to get some perspective when people are discussing texturing. I understood the basic idea of a texture being wrapped around a mesh, but it's nice to get a little more detail.
Just an expansion on my question ... When you say Reach uses one "layer" for base, specular, normal, how exactly does that work? You have a texture of a specified resolution and compression that uses x number of bits for colour, x number of bits for specular, y number of bits normal per pixel? And in the Killzone 3 approach, they are all different layers, meaning different possibly different resolution, precision and compression? I guess I'm asking what the advantage of doing everything in one layer is, when you could split them into different layers for flexibility. Obviously, I'm a layman here, so I'm just trying to understand why one dev would go with one approach and not the other. I can understand having a different texture for the face, where you might want to have higher detail than other parts of the body. That is pretty obvious.
Edit: I guess I'm curious about the semantics of the word layer, and how it's being used here. Is it a pairing of a base colour texture with a normal etc, or is it related to the way the information is stored, or both?
Just an expansion on my question ... When you say Reach uses one "layer" for base, specular, normal, how exactly does that work? You have a texture of a specified resolution and compression that uses x number of bits for colour, x number of bits for specular, y number of bits normal per pixel? And in the Killzone 3 approach, they are all different layers, meaning different possibly different resolution, precision and compression? I guess I'm asking what the advantage of doing everything in one layer is, when you could split them into different layers for flexibility. Obviously, I'm a layman here, so I'm just trying to understand why one dev would go with one approach and not the other. I can understand having a different texture for the face, where you might want to have higher detail than other parts of the body. That is pretty obvious.
Edit: I guess I'm curious about the semantics of the word layer, and how it's being used here. Is it a pairing of a base colour texture with a normal etc, or is it related to the way the information is stored, or both?
Last edited by a moderator:
L. Scofield
Veteran
You can theoretically use completely different resolutions for each texture. We tend to call them channels but anything goes. Mass Effect 2 for example conserves texture memory by using smaller resolution normal maps for almost all characters. You don't really notice it because the color and spec layers have the higher detail, but a precise comparison with the PC version could reveal some subtle differences.
By layers I mean using not a single texture map for one channel (color/normal/spec/glow) but combining together multiple textures. Pixel shaders allow you to perform basic Photoshop-like blending between the various layers - adding, masking, overlaying etc. Most engines have shader editors to visualize it in a WYSIWYG environment with a simple graphic interface for the artists. Obviously you'll get the most out of this method if you combine various kinds of textures.
It is commonly used to cover terrain, where you may use a single, non-repeating 2K or 4K texture (can be a satellite photo or something) to get a major color map, then use repeating, tiled textures for sand, grass, rocks, and another non-repeating map as a mask to blend between them (R is for sand, G is for grass B is for rock, for example). It's a pretty common and old technique, starting at around UE2's introduction or so.
The reason it's not as common on characters is that the details are usually far more unique. You want to precisely paint dents and scratches and dirt and self-shadows and such. So most devs will simply choose a high enough texture resolution (2K per character or so) and make sure to scale down parts that are less visible (legs), then place their cameras so that they don't reveal pixelation.
Still, it's evidently a good option for soldier uniforms, at least on the PS3
Images posted by users of these forums showed much better IQ, and I dont see any point of using the site's screens as references when I have the game
Maybe you should read again an notice the "middleground' comment. It's clear it is not worst case.
