Characters Textures: Next Gen samples and sizes?
polycounter lvl 18
::LONG LONG LONG:::
Now before I even get to attempting to render and paint my current project, I need to get these figured out ASAP.
1. UVW layout and diffuse size. Ut2k4 had the head textures separated from the body at 1024X1024 each. Should I just double the size of this for 2k7 (goal engine of project)? I ask because first thing, team colors were defined on the second map for the body for 2k4. While looking at 2k7 character imagery, Im convinced they are using more than just two 2048X2048 maps for the character to get that level of detail shown.
I was thinking of this split
2048 for head
2048 for all limbs
2048 for clothing and shoes
512 for alpha portions such as hair bangs, eyelashes, shoelaces.
Oh how I wish epic would give us clues to what direction they are going for ut2k7.
Im going to try to have the textures and shaders for:
normal
diffuse
sub surface
ambient occulsion
glow
color spec
(light map needed also?)
Other than the normal map and diffuse, I assume I can drop the resolution by half on the above maps since they wont be as poignant?
Can anyone point to the following tutorials.
A complete next gen or pre render tut that has near all of the above. Especially with skin and sss. Im looking for reasons here, not just "this is what I did", but why they choose to. What lighting, scientific, or artistic theory drove them to such a decision.
Im particularly interested in seeing tonal range to mimick skin via underpainting with greens and blues (or whatever colors and reasons) and building up versus starting with a skin tone.
Sub surface scattering map.. should it show blurred lighter and darker areas where blood veins are? What colors are involved? Or is that just sub surface versus sub surface scattering? Do I need an alpha defining the intensity of such, and where if I do does sss play more or less on the surface of a body?
One that shows how the glow and >colored< spec effect one another (seeing it in max versus seeing it in a real time engine are two different things). Again, especially with skin and cloth spec as per what colors to use.
In some ways, looking at a damn good finished model and shader/texture that I can open in a 3d program helps more than just someones screen shots of process (I'll take what I can get though).
If nothing else, any books that I should look for?
LOL! I can see why now I have such problems with online tuts.. I was always that kid going "why? why?". Something classrooms allow.
::
Then there is a whole idea of parralax maps. From the few vids I have seen of ut2k7 and Gears of War, it almost looks like it could be used on the characters as well. Hell, Im not even sure If I will need to make body parts of this character.
Now before I even get to attempting to render and paint my current project, I need to get these figured out ASAP.
1. UVW layout and diffuse size. Ut2k4 had the head textures separated from the body at 1024X1024 each. Should I just double the size of this for 2k7 (goal engine of project)? I ask because first thing, team colors were defined on the second map for the body for 2k4. While looking at 2k7 character imagery, Im convinced they are using more than just two 2048X2048 maps for the character to get that level of detail shown.
I was thinking of this split
2048 for head
2048 for all limbs
2048 for clothing and shoes
512 for alpha portions such as hair bangs, eyelashes, shoelaces.
Oh how I wish epic would give us clues to what direction they are going for ut2k7.
Im going to try to have the textures and shaders for:
normal
diffuse
sub surface
ambient occulsion
glow
color spec
(light map needed also?)
Other than the normal map and diffuse, I assume I can drop the resolution by half on the above maps since they wont be as poignant?
Can anyone point to the following tutorials.
A complete next gen or pre render tut that has near all of the above. Especially with skin and sss. Im looking for reasons here, not just "this is what I did", but why they choose to. What lighting, scientific, or artistic theory drove them to such a decision.
Im particularly interested in seeing tonal range to mimick skin via underpainting with greens and blues (or whatever colors and reasons) and building up versus starting with a skin tone.
Sub surface scattering map.. should it show blurred lighter and darker areas where blood veins are? What colors are involved? Or is that just sub surface versus sub surface scattering? Do I need an alpha defining the intensity of such, and where if I do does sss play more or less on the surface of a body?
One that shows how the glow and >colored< spec effect one another (seeing it in max versus seeing it in a real time engine are two different things). Again, especially with skin and cloth spec as per what colors to use.
In some ways, looking at a damn good finished model and shader/texture that I can open in a 3d program helps more than just someones screen shots of process (I'll take what I can get though).
If nothing else, any books that I should look for?
LOL! I can see why now I have such problems with online tuts.. I was always that kid going "why? why?". Something classrooms allow.
::
Then there is a whole idea of parralax maps. From the few vids I have seen of ut2k7 and Gears of War, it almost looks like it could be used on the characters as well. Hell, Im not even sure If I will need to make body parts of this character.
Replies
You can just have a look at some examples you like and use the same specs ; this, or use an existing game and render your art in its engine, which will give you obvious limitations and restrictions to follow.
Or, pick up some realtime shaders for max (or modelviewers) and work within their limitations just like you would do with a game engine.
All games have different specificities anyways!
3 2048s? That very excessive. At most you would want a 1024 for face, and 2048 for body. You would never notice a 2048 on a face ingame unless you're doing extreme closeups. I would write more but i just woke up and got in to work, blah
I don't think the UT2K7 nneds 2k textures myself - remember that the biggest res MOST folk run at is 1600X1200 - so painting a head at 2k means the texture is bigger than the screen can handle with noting else on it.
too much theory
just make art.
Stick to just using Diffuse, Specular, Normal (maybe Heightmap too, which can often be multiplied on top of the normal by the engine) and Glow (if you have any parts that need to glow).
Light map is not needed. Ambient occlusion should be baked into the diffuse and specular maps subtly.
Specular map should be in colour. Just one map though, you don't need a greyscale too.
Absolutely don't bother with sub-surface scattering, no need for it at this stage, and it'll just create more work for you, without affecting your chances of being hired.
As EQ has said, your map sizes sound totally overkill. I never understood why Epic used 1024x1024 textures for heads in UT2k3 and UT2k4, you NEVER see that detail in game, even when you're right up close.
I'd say a single 2048x2048 for the whole body and head and clothes sounds fine, plus a 512x512 for any opacity-mapped stuff like hair, as you said.
Don't spend ages trying to make 10 different texture maps that most people will never notice, just concentrate on making the main (most important) 3 or 4 really good.
Diffuse, spec, normal, opacity if necessary and glow. Don't bother with any other stuff, half of it is unnecessary at this stage anyway.
Focus on getting a quality model finished, without fussing over precise technicalities at this stage. Most of that stuff varies per engine and studio anyway, so there's no point learning a particular method since you're likely to have to learn a new, specific workflow wherever you end up.
i need to know how to make one. thanks
That is beyond current gen hardware capabilities, or next gen capabilities, or even next-next-next gen.
Consider the screen space that the object will take up on screen. As a decent shooter player, I've never actually noticed a player models face unless he's lying there dead, and I'm bored and poke around a little.
Head for a shooter need not be over 1024. Full body, maybe 2048, or several 1024s/512s. Spec maps at full rez are a must. Ditto for normal maps. Ambient occlusion can go in the diffuse. Spec color could be lower rez if needed.
Really, that's about all that you need for the most vital stuff. Don't get caught up in being overly technical, as it can actually be a hindrance in just making good art. Keep it simple, and add only the stuff that you really need.
Is it me or does it seem there is some sss going on in the face? Or is that just the colored lighting?
To be honest, I play shooters almost exclusively and feel it's a total waste in putting so much resources into a character's face, when I will cap them in .4 seconds at 30-200 yards away. 2048s on the face? How would I ever know that, except by the crappy framerate?
I realize I most likely will have this done before ut2k7 is out, but I wanted to have at least a sdk so people could get it in (might have a animator I know help with some animations), and make it the less fuss possible for them to do this. If I have to rebake everything to different maps, and worse some areas need larger areas then before...
Plus, I still have no idea what shade of blue and red 7 will use for teams.. And will that have glowing effects for those teams?
Would you guys estimate by the above image whats going on? 2 2048s? 1 2048 and 1 1024? 3 1024s (head, upper, lower)?
Just UV-map intelligently, make the best use of the pixels that you can. Don't go overboard on image sizes just cos it seems like other people are.
Like others in this thread have said, just go for making the best possible artwork. You know the rough technical specs (the quake4 art reference on iddevnet.com is very useful for this), so just get to finishing a full model to current gen.
There's no point worrying about what UT2k7's absolute final specs are going to be at this stage.
BTW i dont think UT2k7 actually has much of a facial animation, look at the models. They all seem to have some kind of expression on their face but its too detailed to be dynamic unless they morph normal maps and crap but it would be a waste for online shooter.
http://video.google.com/videoplay?docid=1812427489796949679
going. Have you been watching the trailers/
Maybe I wont bother for the in game play at this time versu just in engine. So if I wasn't, how exactly would you split the above texture sizes if I wanted to mimick that detail? I really want to learn these techs. Also, is there sss going on or not from what you see? Facial morphing and/or facial boning was one of the goals with this project.
See? This is a realtime character with a facial rig.. I agree with you all though its not the in game play version. It however is in engine.
The larger the texture size, the more detail I can add in. Im one of those "God is in the details' people.
Oh god, am I becoming a cinematic realtime character artist unknowingly? Geezuz, I keep pushing myself further and further into corners.
Toomas, it has some sort of pretty advanced facial thing going. Have you been watching the trailers
[/ QUOTE ]
All the shots i have seen has the characters with the same expressions.
If by "trailers" you mean the UE3 tech demos with UT2k7 assets (probably with bloated specs) then yes they have facial animation thingy going.
I find its best to follow my old mothers phillosophy in art, as in life - Just because you can do something (whether using vast amounts of texture space or drinking until you lapse into a coma), doesn't mean you should.
Get something that looks good overall, don't worry about every little pore and wrinkle.
Do not worry about them at this stage. There is no point trying to make something that there is very little info on yet unless you're a developer for UE3.
Plus, they make great screen and render shots for portfolios.
I really still want to investigate sss or ltm. I saw a technical artist from GPG demo it once in nvidias shader program, and I really thought it was f'ing cool. If I can remember his email, maybe he could give me some pointers on this part.
So I guess the only real parts that I need some good books or online tutorials would be working with colored specs (only have used greyscale ones thus far) and what colors are used for light interplay on different surfaces, and creating damn good/fine skin diffuses. I guess seeing if Antropus (CG character god IMOHO), has any of this would be a good start.
Thanks for helping me get a clearer image all.
Im going to have some questions more advanced boning/rigging questions particualarly with efficiency, facial options, helper, and spring type of bones pretty soon. Expect more Emo at that time.
Engine/video card mipmapping will reduce a lot of your work to optimize rendering of the texture.
Remember, most (if not all) game engines are geared towards optimisation. It wants to render things as efficiently as possible. 3d applications are skewed towards visual quality. What you see in your 3d app won't be the same as what you see in game.
Don't get hung up on details. Be practical and think about how object is going to be used/seen. At work, we have a way of measuring the mip map levels on a per polygon basis. Before I even texture anything, I dump it into the game with the wirefram texture slapped on. I then look into how many times the texture is being mip-mapped in our game camera. If the texture is getting mipped once, then that means my texture is twice as big as needs to be.
Not only does providing the appropriate sized texture for the game efficient (takes less to render), it's also efficient for me as I know exactly what resolution to paint at.
Even an FPS game has a set distance on how far back the camera is to any given object.
Also, higher resolution isn't the only way to achieve detail. There are all sorts of optical illusions you can do to trick the eye into thinking there's more than what's actually there.
(man reducing the polycount is hard with characters who need normal maps without f'ing it up. My respect has gone up 10 fold for those that must do this everyday since starting this portion. Im still over 10000, and im worried about not following the high poly close enough)
Any "interior" detail that will rarely be seen in silhouette can be really low-poly. Check out Pior's stuff like Aquaboy and his new Q4 Razor model for how to make a nice lowpoly.
Lets Get Small
Understanding MIP Mapping
Kevin Bjorke, NVIDIA
June 2005
I thought it did but it doesn't mention that specific trick. He does talk about an fx shader for it though. And he talks thru some other cool ideas + explains some of the mip/resolution issues people have been talking about here. Worth a read.