Pro Spec and AO Map Creation tips [req]
polycounter lvl 17
Can I get some pro tips on creating spec maps for games?
I've heard of people creating them from normal maps as well as from diffuse maps. I'm unsure what is the best method. Also, What areas in general are best to be brighter and darker?
Also, AO maps, i've found that they add a strange look in Maya and other engines. Can someone explain the best way to make use of AO maps in games?
Thanks!
I've heard of people creating them from normal maps as well as from diffuse maps. I'm unsure what is the best method. Also, What areas in general are best to be brighter and darker?
Also, AO maps, i've found that they add a strange look in Maya and other engines. Can someone explain the best way to make use of AO maps in games?
Thanks!
Replies
It's also useful for seeing where your high poly's details are on the unwrap so you can paint the correct colours on.
On the left is flat grey multiplied by a Max AO map with a Max normal map, AO directly connected to specular, dynamically lit in UDK.
On the right is flat grey with a Max normal map and a white specular map.
Having a proper AO map somewhere in the mix just makes the finished product look more like a solid object and less like a pile of polygons.
http://eat3d.com/texturing
Almost every single engine has a 'lite' version of AO as James said, they account for distance between objects on a small scale, and only the polygons, not the micro detail of textures themselves, hence why many models can look flat without abit of shading done by hand.
With that said, I honestly feel like AO simply doesn't cut it anymore, simply pasting it in your Diffuse and calling it a day. It doesn't make sense (especially for, if I read you correctly, faces and other organic objects) where they scatter the light, adding in a static AO will just make it look really tacky in more cases then one.
I instead bake out the Cavity Map, since they don't broadly darken your textures, and instead darken the really tight small details, which play really well with AO in most engines as they're not fighting light information for the human eye, especially on light scatter friendly objects.
If you really want to get fancy, you can do what the guys at Naughty-Dog did, which is anything that is in the shadows shows the AO bake ontop of your diffuse, and anything in the light doesn't.
As for making them, I suggest baking out your Cavity/AO maps, if you're hand painting them, then you're already putting in said detail.
As for Specs, many people just greyscale their Diffuse and call it a day, I much prefer using my Cavity and this process here: http://www.modwiki.net/wiki/Start_a_Specular_map_with_a_Normal_map to generate my Spec Maps in a way that makes sense.
It definitely helps to bake it out and then use it as appropriate for your taste and engine of choice.
What I do is that I copy my entire diffuse group in Photoshop, greyscale it (Hue Saturation layer) and then I work layer by layer, adjusting everything - removing and adding layers, using a lot of levels-layers, HueSat-layers, Curves etc, changing blend modes for wear and tear, dirt, and so on. Also important: color. You can't make a material such as gold without a yellow spec and a brown diffuse (unless ofc you are going for a cartoony style and just paint highs and lows onto the diffuse).
Ofc there are many ways of doing a spec, but "just" greyscaling is - like I put it: stupid.
Oh god.
OH GOD NO.
You can do this easily in crazybump by pasting in your normals and upping the "enhance detail" slider, what I generally use for a "cavity map". Though this just helps to add a bit of pop, or is used for creating selection masks for various effects, I would never call it "generating a spec map".
Also... hand painting ao and generating spec maps, lol wut.
More or less this, but remember that diffuse, spec and normals all work together. Spec isn't something you tag on at the end but something you should be working on the entire time. Not sure if thats what you mean't by copy over from your diffuse or not(ie: copy when you diffuse is "done").
As far as AO:
Highpoly exploded bake multiplied
+
Lowpoly combined bake(with bits that need to animate removed) multiplied
+
CB cavity map overlayed at 50%
over diffuse and spec.
(check brink assets in my signature for what this looks like, the lowpoly untextured shots are just AO).
I don't start working on my spec until I'm about halfway done with the diffuse though - but that's just a matter of taste I guess...
I can't see your signature btw - you know why?
Likewise, using a greyscale specular map (1 channel) is also sometimes the best way to do things. It's faster because you don't have to worry about adjusting colors, and it's cheaper because 1 channel takes less memory than 3 channels.
The best looking spec map will always be 3 channels, made with lots of love, and with a good gloss map to back it up. But not every asset calls for that much work and that many channels of texture memory.
As for workflow, I actually like to make my diffuse map completely (while thinking about spec, of course), then while I am working on my spec map I will go back and tweak things in the diffuse to compensate.
Oh, and a little trick I've started using is multiplying sections of my diffuse on top of my spec (adjusting the spec strength to compensate, of course). This can be done easily in a material editor, using a 1 bit mask. It can let you get important color information into the spec without requiring an entire 3 channel spec map. Of course it does mean more material functions. Everything is a tradeoff.
Welcome [your SN]
User CP
Settings & Options, Edit Options
Thread Display Options, Visible Post Elements
[x]Show Signatures (Check me!):poly136:
never rationalize poor workflows with "time constraints". it takes minutes to make a spec map that looks infinitely better than a diffuse desaturation, even if it's not the best it will be better than that.
grayscale in alpha does not take less texture memory due to how dxt compression works. in fact, two full color dxt1 textures are the same size as dxt5 with grayscale alpha. http://www.opengl.org/wiki/S3_Texture_Compression
as eq mentioned, all the maps work together--thus working on all the maps simultaneously is the fastest and most logical way to texture.
ive been awake too long
A 1 channel map can be greypacked with other 1 channel maps, so that multiple textures can be put into a single image file. For instance, I can have one 3 channel targa that stores specular (red channel), transparency (green channel), and emissive (blue channel). So yeah, a greyscale spec map can certainly save memory, as long as I pack it in with something else. You can even put a greyscale texture in the blue channel of your normal map (though there are reasons not to). This is ignoring different compression settings such as TC_Greyscale (in udk), which are designed for making greyscale images look better while using less memory.
My reason for making my diffuse fully before working on my spec is that I can duplicate my diffuse folder, rename it spec, and start adjusting layers without spending any time making sure my diffuse and specular match each other. Otherwise I have to be modifying two different sets of layers, which are often nearly identical. For instance if I were altering a dirt layer for both spec and diffuse at the same time, I would need to alter one, then make a selection of that layer and alter the other based off of that selection. Doesn't sound like much, but it adds a lot of time when you wind up tweaking dozens of layers, dozens of times.
yes, i've worked with "packed" grayscale textures on a few titles and every time i wished i just had colored specular and no gloss at more than one point during the projects. while it's true you can save vram with a "packed" texture, you need to put in more work to separate materials and give them color. many accent materials become impossible to represent properly without colored spec as well, and pretty dynamic color variation goes out the window.
oh, also that's not to say it CAN'T look good, just that it takes more effort to look good and all your color has to be on the diffuse.
i used to be of the same mindset as you--make diffuse first and duplicate/tweak for fast specular and gloss. then i started blocking out materials first and it made a LOT more sense to work on them at the same time. i know my/eq/per's spec and gloss have lots of unique layers on them and look radically different from the diffuse, creating a material that's very light reactive.
you tend to use a lot less layers when you're working like that, so creating all the maps simultaneously gives you a more dynamic result in less time. spending all that time trying to make just your diffuse look good on its own then shoehorning in a spec map after all that work requires more time and more layers to manage for ultimately a less reactive result.
So, in order of best quality to most efficient(yes the various methods are inversely proportional).
1. Full color spec with gloss in alpha channel, 32 bit image, compressed is twice as larger as #2
2. Full color spec, gloss set via material, 24 bit image
3*. Spec, Gloss and something else (Emmisive**, Alpha***, etc) packed into a single 24 bit image, same size as #2
4. Specular derivative from diffuse map, usually with some contrast adjustments etc via shader. Essentially free memory wise. Can be useful for extremely simple textures where you're only representing 1 material type per image
* Only viable if your asset actually requires 3 additional maps
** Emmisive can often be sized down 1 or 2 notches, saving 4x and 16x what an additional full color texture would cost. Color Emmisive can be useful as well
*** Alpha generally shouldn't be applied to full assets for performance reasons, but instead a smaller secondary material with a smaller texture. You don't want to run alpha sorting on an asset that is 95% opaque, only on the section that actually needs alpha.