seesm to me there there is not much difference visually betwen using a numerical value of 10 for a spec pow (for a skin shader) and and specpower map.
seems a like a waste of resources to use the extra texture in this case
You use a map when you want to change the spec power across a surface - say for shiny buttons (tight, hot spec hits) on a jacket (soft, wide spec hits).
yes and sometimes on the high poly model since all the different parts can use completely seperate materials sometimes you won't even need certain texture maps at all for some of things but then for the low poly since everything has to basically boiled down to 1 material usually then you're probably gonna want to have all those textures available in the engine.
but you would n't texture the high res model and would not need to do separate materials for each element. normally you would just bake the detail in the normal map and texture using the normal map as a guide.
normally you just get 1 texture map for the fleshy bits and 1 for the body/clothy bits etc .
I suppose in an ideal world you would have specialised shader for each type of object be it metal cloth etc, but extra material id's are expensive so it ain't gonna happen any time soon
The glossiness map is very important for organic models. If you work with hi-polygons models for cinematic you will see it.
Just look over all the games using normal maps now, the result is very plastic and metallic, very wet, really ugly, horrible. I recently played Prey and that crappy look seems to be almost inevitable (Oh jen!).
With the actual hardware, to be bothered with the budgets is really silly, imho. There are too many games in development using complex shaders .fx. If you only can use specular maps and a map for auto-iluminations, due to the budgets, there are not too many problems; you can do it very good aswell, can't you?
What is another map? more work? is that?. I really prefer to work on 2 maps only (normal map is generated, so the work is less tedious), but if i want to achieve a result, i need to use more maps than a simple specular. Ok, it's more work, but with some tweaks on the bump map you get specular and gloss.
[ QUOTE ]
With the actual hardware, to be bothered with the budgets is really silly, imho
[/ QUOTE ]
Well, sorry to say it, budgets do matter. EA might have superb 20,000 polygon boxers in Fightnight 3 with loads of sexy shaders, but thats 2 men in a room. Unreal, GOW, Rainbox 6, Tony Hawks - those games using big streaming environments simply CAN'T DO THAT.
Its not much more work for the artist, but it is more work for the game. Another pass. You want as few as possible.
Bad normal maps in games are due to bad specular settings used in conjuntions with them. I've had the problem myself - it was the spec falloff that was making everything too harsh. Working with the spec map got me much better results. THe spec and normal look better that just diffuse. I could use cloth sims and SSS on the skin, and hair effects and that'd look better - but when you've got 20 characters on screen projecting shadows onto an environment and then using some sort of AI and the game itself running everything grinds to a halt.
Budgets will always be there, and i'm agree with you, they matter a lot. But just wanted to say that with the hardware we have (not all ppl) we have better and better budgets to get nicer things. We should not be always with the same budgets (1500-3000 polygons and 1024 textures). That's all.
To know how work the SSS shaders and all the effects is essential for the next gen artist. People who don't know to use them correctly make bad graphics. Just look over all the cg forums, only 2-4 guys do nice shaders and nice things with it. Is always the same, if you know how to do the things good with only a specular map, you will do better things using more maps. there are also guys who do nicer models using only a color map, they don't need a normal map and of course they hate them. In the japanese videogame industry they have opted for more polygons instead of expensive shaders for the cpu. They know the limits of the hardware very well. I really rather prefer a game without normal maps than with. That "black" aspect is unevitable, you feel forced to up the gamma to 3 or 4.
How many maps we use in characters?
normal, bump, spec, ilumination, masks... where's the problem to add a gloss map if we need it?. If someone want to have a less amount of memory for textures, just use 256 color palettes or grayscale for the maps like spec and not those tgas for all textures. Optimize the graphics. there are too many times a texture works better with a 256 color palette or less, of course, depending of the amount of colors on the image (for example one a bit monocromatic).
The evolution is more maps and more polygons, more detail. In a few years you will se how models use displacement and adaptative subdivision in realtime plus all those complex shaders. No doubt, more work for us and more time to do a character, and the worse... more time to finish games and better and more qualified profesionals will be needed.
Let's see how works games such as gears of war with those huge amount of 2048 maps, or better, the new game from crytek, Crisis.
Anyways, I was talking about texture budgets. Talking about the amount of polygons mmm.. you should play Company of heroes, or at least take a look over it. I have never seen a game with such amount of texture detail and polygons for a RTS. Another example of detailed models are in Oblivion, i have played with more than 12 guards on screen. Resident evil Game Cube version is other good example. The actual hardware is very powerful, having a 6800gt you run games smoothly, at a 40+ fps on a 3ghz intel prescot. Game companies need to take better advantage of the hardware. I really don't know why the better games for a videoconsole appear when the videoconsole has its days counted...
When i played oblivion first time it ran at a 24-32fps, and now with the new nvidia driver+game patch the game runs at more than 60fps in the same system with all graphic settings to the max, 1280x1024 resolution. I must think they don't take good advantage of the hardware, or they must have bad programmers, dunno.
20000 polygons are nothing in those days. Actual Hardware can handle more than 3kk polygons if you have a decent machine.
[ QUOTE ]
I really rather prefer a game without normal maps than with. That "black" aspect is unevitable, you feel forced to up the gamma
....
I must think they don't take good advantage of the hardware, or they must have bad programmers, dunno.
[/ QUOTE ]
Black can be avoided simply by using more tht one light, or by adding a diffuse cubemap pass (big improvement there).
New hardware always takes a long time to come to grips with, they never expose everything right off the bat. Besides the dev tools are always minimal to start, until the console's been around a while.
<edit> Here's a super-old example of the cubemap method we're using.
Lower-left is just the single spot light alone, lower-right is the cube alone. ATI has a great util for making these cubes too.
I'm not sure how to explain this and how it would relate to the game industry but for prerendered work having access to a gloss map is very helpful. It's also kind of hard to explain on how to use it since you can do so much with it when combined with a color, spec color, specular map and bump. Sometimes you don't want the spec color and spec map level to be the same. The gloss map can kind of be looked at as a mask the show the visibility of the spec and also controls it's shape. This can be an animated map to say create running water easily over a surface.
This might be a bad example but here it goes.
Oh this does not follow how human skin reacts... Imagine you have a face and for some areas you need the spec color to be bright blue and others dark blue, since that is how that skin reacts when it is lit the same amount of light. Futher more this skin would have very high amounts of spec level in the dark areas and sharp yet dim levels in the bright area of the spec color. To do that you would use the gloss map to describe these areas. For example you could have a high contrast area in the gloss map for the dim sharp spec values.
I see the gloss map as something that subracts the power of spec map in a material. Sometimes the brightness of a color in a spec color maps doesn't match it's intensity. I'm not sure this explanation makes sense, you kind of have try it to see the difference.
I'm not sure how spec maps are used in games.
Spec color = I use it as a plain color to describe how the specular color is supposed to to look when lit.
Spec level = grayscale image used to control the intesity of spec color
Gloss - can be used to describe the shape of the highlights, where they are supposed to be sharp or soft. I see if as kind of a mask for the spec aspect of a material if that makes sense.
I always thought spec falloff was the gloss map as long as it could be defined with an image.
Most of us don't make games for PCs - most of us make games for consoles.
All of us who make games don't make games that only feature characters - they have enviroments and vehicles. We can't piss away all the frames on shaders for the characters. We can't piss away 10,000 polygons on each character when we have a city to render in the background.
[ QUOTE ]
To know how work the SSS shaders and all the effects is essential for the next gen artist. People who don't know to use them correctly make bad graphics.
[/ QUOTE ]
Absolute fucking cock. There are plenty of people out there making stunning game art who don't the first thing about those shaders.
It dosen't mean that all these need to be ported to an engine just for the sake of being nextgen, does it.
Alex, if you need to render human skin in max you certainly need a gloss map (I never did that hence I simply trust you on this). For a game one simply does what he can within the budjet that has been defined earlier in the production cycle - and for such things as skin, the general approach is to get a specific realtime shader achieving the specific look of the desired material (with maps as inputs, or not).
HANDPAINTED FTW! (and I'm not joking here : an artist able to paint a convincing portrayal of a character either in natural media or in PS is very likely to succeed more than someone who only knows his way throuhg interconnected texture maps, when it comes to produce a realtime or prerendered model of the same subject)
no...what this is about is that if you can create something pleasing with only two shaders why toss in loads more just to eat up memory ? just because they can be done ?
hm..i think ill do relief/parallax/normal , with diffuse, sss, gloss , specular , second specular , fallof, cha cha maps...
[ QUOTE ]
That "black" aspect is unevitable, you feel forced to up the gamma to 3 or 4.
[/ QUOTE ]
Entirely incorrect, as EricChadwick pointed out, ambient cubemaps are now used to fake bounced environment light and keep the models from shading too dark. There are other lighting solutions for this too. Possibly you're too stuck in the past thinking only of the Doom3 and Quake4 lighting implementations.
[ QUOTE ]
Let's see how works games such as gears of war with those huge amount of 2048 maps
[/ QUOTE ]
Ah, so you've seen the flat textures and environment texture specs for Gears of War? You've got the image dimensions and asset polygon counts there? Maybe you'll be surprised when the game comes out and you can check their assets for real instead of blind conjecture.
Know the ins-and-outs of the engine you're working on and what it's capable of and you'll be fine. Information to successfully do your job is things like:
-budget counts (tri per character, per world asset, per overall frame, etc.)
-rendering capabilities
-shader control creation (now generally done using a GUI but sometimes done using, sigh, notepad)
-specularity control (what kind of spec is it, how much value is given using control maps, etc.).
-etc.
A great example of knowing what you know, then applying it to your position is what I've been through working with the Source engine for the past year. Environment map masks - Source's current specularity pass - are a touchy bitch. Values and distribution within a height map that I was normally use to pre-Source were waaaaaaaay outta control once applied TO source. Now that I know the ins-and-outs of how it displays them its easy enough to get that level of specularity I'm looking for. Make sense?
You don't need to know SSS, or parallax mapping or any of that other shit unless its directly related to your position within the company and how these features relate to do your job properly.
'Blaizer', your expression of your 'point of view' was annoying since you're reeling off these facts to us which clearly lack experience. Know your audience. Most people in this thread are videogame developers so lose the patronising 'this is how it is' tone. I dont know why the fuck polycount is so chock full of cocky noobs lately but its reminded me why I dont hang out here much anymore.
As Rick already tried to explain to you, on a production pipeline the fact of the matter is that you never *EVER* have the resources to play with that you might think you have. And yes, that includes cutting edge hardware and engines. By the time there is a game running in the background with AI and physics and other code going on and you have to get everything into RAM, you simply HAVE to make cutbacks, choices and sacrifices. SSS? wtf, maybe in in game cinematics for one or two characters. When the fuck else would you even see and or appreciate it? Bang for the buck, thats what videogame Art is all about. Intelligent choices. All the games you cite as examples have performance problems on even the most cutting edge hardware.
And I've no idea why you bring up the uncanny valley phenomenon in response to Erics lighting example. Look up what uncanny valley actually means.
Replies
seems a like a waste of resources to use the extra texture in this case
i brought it up coz i was getting the impression it was not common and i felt Doom3 had influenced that.
normally you just get 1 texture map for the fleshy bits and 1 for the body/clothy bits etc .
I suppose in an ideal world you would have specialised shader for each type of object be it metal cloth etc, but extra material id's are expensive so it ain't gonna happen any time soon
Just look over all the games using normal maps now, the result is very plastic and metallic, very wet, really ugly, horrible. I recently played Prey and that crappy look seems to be almost inevitable (Oh jen!).
With the actual hardware, to be bothered with the budgets is really silly, imho. There are too many games in development using complex shaders .fx. If you only can use specular maps and a map for auto-iluminations, due to the budgets, there are not too many problems; you can do it very good aswell, can't you?
What is another map? more work? is that?. I really prefer to work on 2 maps only (normal map is generated, so the work is less tedious), but if i want to achieve a result, i need to use more maps than a simple specular. Ok, it's more work, but with some tweaks on the bump map you get specular and gloss.
Think hi-poly, go hi-poly. Games are evolving.
With the actual hardware, to be bothered with the budgets is really silly, imho
[/ QUOTE ]
Well, sorry to say it, budgets do matter. EA might have superb 20,000 polygon boxers in Fightnight 3 with loads of sexy shaders, but thats 2 men in a room. Unreal, GOW, Rainbox 6, Tony Hawks - those games using big streaming environments simply CAN'T DO THAT.
Its not much more work for the artist, but it is more work for the game. Another pass. You want as few as possible.
Bad normal maps in games are due to bad specular settings used in conjuntions with them. I've had the problem myself - it was the spec falloff that was making everything too harsh. Working with the spec map got me much better results. THe spec and normal look better that just diffuse. I could use cloth sims and SSS on the skin, and hair effects and that'd look better - but when you've got 20 characters on screen projecting shadows onto an environment and then using some sort of AI and the game itself running everything grinds to a halt.
To know how work the SSS shaders and all the effects is essential for the next gen artist. People who don't know to use them correctly make bad graphics. Just look over all the cg forums, only 2-4 guys do nice shaders and nice things with it. Is always the same, if you know how to do the things good with only a specular map, you will do better things using more maps. there are also guys who do nicer models using only a color map, they don't need a normal map and of course they hate them. In the japanese videogame industry they have opted for more polygons instead of expensive shaders for the cpu. They know the limits of the hardware very well. I really rather prefer a game without normal maps than with. That "black" aspect is unevitable, you feel forced to up the gamma to 3 or 4.
How many maps we use in characters?
normal, bump, spec, ilumination, masks... where's the problem to add a gloss map if we need it?. If someone want to have a less amount of memory for textures, just use 256 color palettes or grayscale for the maps like spec and not those tgas for all textures. Optimize the graphics. there are too many times a texture works better with a 256 color palette or less, of course, depending of the amount of colors on the image (for example one a bit monocromatic).
The evolution is more maps and more polygons, more detail. In a few years you will se how models use displacement and adaptative subdivision in realtime plus all those complex shaders. No doubt, more work for us and more time to do a character, and the worse... more time to finish games and better and more qualified profesionals will be needed.
Let's see how works games such as gears of war with those huge amount of 2048 maps, or better, the new game from crytek, Crisis.
Anyways, I was talking about texture budgets. Talking about the amount of polygons mmm.. you should play Company of heroes, or at least take a look over it. I have never seen a game with such amount of texture detail and polygons for a RTS. Another example of detailed models are in Oblivion, i have played with more than 12 guards on screen. Resident evil Game Cube version is other good example. The actual hardware is very powerful, having a 6800gt you run games smoothly, at a 40+ fps on a 3ghz intel prescot. Game companies need to take better advantage of the hardware. I really don't know why the better games for a videoconsole appear when the videoconsole has its days counted...
When i played oblivion first time it ran at a 24-32fps, and now with the new nvidia driver+game patch the game runs at more than 60fps in the same system with all graphic settings to the max, 1280x1024 resolution. I must think they don't take good advantage of the hardware, or they must have bad programmers, dunno.
20000 polygons are nothing in those days. Actual Hardware can handle more than 3kk polygons if you have a decent machine.
take a look over this website, im sure you will like it
http://www.illusion.jp/index2.html
an example of the game:
It's hard to see a blocky model... And if you say that there's only a girl, they have a test with a lot of them.
http://www.dl.illusion.jp/download/actio...pelay_benchmark
http://www.illusion.jp/technicalnews/index.html
I really rather prefer a game without normal maps than with. That "black" aspect is unevitable, you feel forced to up the gamma
....
I must think they don't take good advantage of the hardware, or they must have bad programmers, dunno.
[/ QUOTE ]
Black can be avoided simply by using more tht one light, or by adding a diffuse cubemap pass (big improvement there).
New hardware always takes a long time to come to grips with, they never expose everything right off the bat. Besides the dev tools are always minimal to start, until the console's been around a while.
<edit> Here's a super-old example of the cubemap method we're using.
Lower-left is just the single spot light alone, lower-right is the cube alone. ATI has a great util for making these cubes too.
This might be a bad example but here it goes.
Oh this does not follow how human skin reacts... Imagine you have a face and for some areas you need the spec color to be bright blue and others dark blue, since that is how that skin reacts when it is lit the same amount of light. Futher more this skin would have very high amounts of spec level in the dark areas and sharp yet dim levels in the bright area of the spec color. To do that you would use the gloss map to describe these areas. For example you could have a high contrast area in the gloss map for the dim sharp spec values.
I see the gloss map as something that subracts the power of spec map in a material. Sometimes the brightness of a color in a spec color maps doesn't match it's intensity. I'm not sure this explanation makes sense, you kind of have try it to see the difference.
I'm not sure how spec maps are used in games.
Spec color = I use it as a plain color to describe how the specular color is supposed to to look when lit.
Spec level = grayscale image used to control the intesity of spec color
Gloss - can be used to describe the shape of the highlights, where they are supposed to be sharp or soft. I see if as kind of a mask for the spec aspect of a material if that makes sense.
I always thought spec falloff was the gloss map as long as it could be defined with an image.
Any thoughts would be nice, thanks
Alex
All of us who make games don't make games that only feature characters - they have enviroments and vehicles. We can't piss away all the frames on shaders for the characters. We can't piss away 10,000 polygons on each character when we have a city to render in the background.
[ QUOTE ]
To know how work the SSS shaders and all the effects is essential for the next gen artist. People who don't know to use them correctly make bad graphics.
[/ QUOTE ]
Absolute fucking cock. There are plenty of people out there making stunning game art who don't the first thing about those shaders.
Like, if you check the available map slots in max that's what you have:
http://www.cglearn.com/tutorials/max3/11_materials_still_life_pt1_23.jpg
It dosen't mean that all these need to be ported to an engine just for the sake of being nextgen, does it.
Alex, if you need to render human skin in max you certainly need a gloss map (I never did that hence I simply trust you on this). For a game one simply does what he can within the budjet that has been defined earlier in the production cycle - and for such things as skin, the general approach is to get a specific realtime shader achieving the specific look of the desired material (with maps as inputs, or not).
HANDPAINTED FTW! (and I'm not joking here : an artist able to paint a convincing portrayal of a character either in natural media or in PS is very likely to succeed more than someone who only knows his way throuhg interconnected texture maps, when it comes to produce a realtime or prerendered model of the same subject)
Pior, i hate 3dsmax's shaders.
Rick_Stirling, are you saying that 3d artists do not need to understand those particular shaders or all shaders in general?
hm..i think ill do relief/parallax/normal , with diffuse, sss, gloss , specular , second specular , fallof, cha cha maps...
[/sarcasm]
Rick_Stirling, are you saying that 3d artists do not need to understand those particular shaders or all shaders in general?
[/ QUOTE ]
I was responding to the comment inferring that artists who don't know shaders make bad art.
And...I agree with you.
That "black" aspect is unevitable, you feel forced to up the gamma to 3 or 4.
[/ QUOTE ]
Entirely incorrect, as EricChadwick pointed out, ambient cubemaps are now used to fake bounced environment light and keep the models from shading too dark. There are other lighting solutions for this too. Possibly you're too stuck in the past thinking only of the Doom3 and Quake4 lighting implementations.
[ QUOTE ]
Let's see how works games such as gears of war with those huge amount of 2048 maps
[/ QUOTE ]
Ah, so you've seen the flat textures and environment texture specs for Gears of War? You've got the image dimensions and asset polygon counts there? Maybe you'll be surprised when the game comes out and you can check their assets for real instead of blind conjecture.
I said this:
[ QUOTE ]
People who don't know to use them correctly make bad graphics.
[/ QUOTE ]
And Rick says this... really incredible.
[ QUOTE ]
I was responding to the comment inferring that artists who don't know shaders make bad art.
[/ QUOTE ]
Why have you assumed that? Where did i say that? bad art?? i speak english so bad?
Maybe i'm talking nonsenses, maybe is the opposite, well the time will give the reason. Sorry for my point of view.
...the rabbit tweak, it still looks ugly. Uncanney Valley.
I don't know why people insist on calling me as "Blazier". I must assume you need glasses
[ QUOTE ]
Possibly you're too stuck in the past thinking only of the Doom3 and Quake4 lighting implementations.
[/ QUOTE ]
You assume wrong things dude.
i quit this thread
-budget counts (tri per character, per world asset, per overall frame, etc.)
-rendering capabilities
-shader control creation (now generally done using a GUI but sometimes done using, sigh, notepad)
-specularity control (what kind of spec is it, how much value is given using control maps, etc.).
-etc.
A great example of knowing what you know, then applying it to your position is what I've been through working with the Source engine for the past year. Environment map masks - Source's current specularity pass - are a touchy bitch. Values and distribution within a height map that I was normally use to pre-Source were waaaaaaaay outta control once applied TO source. Now that I know the ins-and-outs of how it displays them its easy enough to get that level of specularity I'm looking for. Make sense?
You don't need to know SSS, or parallax mapping or any of that other shit unless its directly related to your position within the company and how these features relate to do your job properly.
As Rick already tried to explain to you, on a production pipeline the fact of the matter is that you never *EVER* have the resources to play with that you might think you have. And yes, that includes cutting edge hardware and engines. By the time there is a game running in the background with AI and physics and other code going on and you have to get everything into RAM, you simply HAVE to make cutbacks, choices and sacrifices. SSS? wtf, maybe in in game cinematics for one or two characters. When the fuck else would you even see and or appreciate it? Bang for the buck, thats what videogame Art is all about. Intelligent choices. All the games you cite as examples have performance problems on even the most cutting edge hardware.
And I've no idea why you bring up the uncanny valley phenomenon in response to Erics lighting example. Look up what uncanny valley actually means.
Oh btw, daz. You are a very fine upstanding member of society.
Btw, this is now a thread about gloss objects.