Melan on 27/10/2009 at 09:04
Thank you, I will try these suggestions. At this point, I am mostly looking at trims, windows and doors to spice up TDM's default selection. CGTextures recently posted a huge number of beautiful textures one of their contributors took in India; getting them to tile is one thing, but I have had no previous experience with normalmaps since Dromed didn't use them.
I will post an example or two when I am home from work.
Judith on 27/10/2009 at 11:28
In the Gimp normalmap plugin use value like 5 or 6 as a base, then stack Gaussian-blurred layers in overlay mode, like in this CGTextures tutorial (e.g. starting from 5, increasing the blur 1 point per new layer, ending on 10 or 15 if necessary). You can do this really quick by pressing Ctrl+Shift+D (duplicate layer) and then Ctrl+Shift+F (show last filter's dialogue window). If this is a tileable texture, remember to expand it before blurring layers, as this will leave seams on its borders.
Melan on 27/10/2009 at 11:32
Do you mean "expand" as in resize, for example by about 5%? Do I convert it back to its original size later?
Judith on 27/10/2009 at 13:23
Quote Posted by Melan
Do you mean "expand" as in resize, for example by about 5%? Do I convert it back to its original size later?
Nope, I mean you'll have to tile it 3 times (e.g. 3072 if you have a 1024 texture), and use the crop tool to select the tile in the middle. Then you'll have to tick the "expand from center" option and increase the crop size a bit, it can be 64 or 128 pixels. When I work with 1024 textures I usually expand it to 1280. After that you can start duplicating the layers and blurring them. When you finish and normalize your normalmap, don't forget to crop it back to 1024.
Renzatic on 27/10/2009 at 15:52
I do things a bit differently, usually starting out with a straight import from the normal plugin, overlay 1 or 2 of them to form a base for your fine details, then start building with a blur factor of 2 from there. If you're finding yourself a bit confused as to exactly what you're doing here, think of it like this. Each time you overlay the same normal on top of itself, you're adding depth, and each gaussian blur adds curvature and defines the larger details. By the time you're done, you might have a stack of 20 or more texture layers.
The one thing I think you should get the hang of first and foremost is making good greyscale heightmaps. This is ultimately what's gonna define your details when you start bringing things up with your blur layers. Sometimes you'll find just running your base diffuse through the normalmapper is good enough. But once you start doing it more, you're gonna find that method is, at best, a random wild turkeyshoot. Sometimes you get a good normal from it with all your details defined exactly how you want, most of the time you don't come anywhere near. Doing a good height lets you control exactly what you bring up.
Alternately, you can also try (
http://www.crazybump.com/) Crazybump. It basically does what we're detailing above, but without as many extra steps and a nice preview window to help you along. I usually use a combination of the blur method and this when making my normals.
The only downside? At $100, it's a bit overly expensive for a random hobbyist just getting into this for the first time. Unless, of course, have a student ID or know someone who does.
edit:
:facepalm: Oh damn man, that wasn't a finished texture. I just posted it up to see if Beleg liked it. Oh well, guess it's up for the grabbing now. :P
Judith on 27/10/2009 at 17:20
Quote:
one thing I think you should get the hang of first and foremost is making good greyscale heightmaps.
That's true, when you're starting your work with normalmap, the first thing you must do is to desaturate your diffuse image, and tweak the heightmap if necessary.
The most common example of a bad normalmap is the brick wall with white mortar. If you just create a normalmap out of diffuse texture, you'll get convex mortar and concave bricks (or at least placed a bit behind it). You have to look at your desaturated image, see whether convex parts are bright grey or white (something above RGB=128 - this is the "absolute grey", a kind of middle value), and the concave parts should be darker grey or even black. Only after that you should apply the normalmap filter and start working with layers.
You can use many different methods here, painting, adjusting levels, sometimes a simple color inversion will be enough.
Renzatic on 27/10/2009 at 17:54
One thing that's really handy in Photoshop is the black and white adjuster. It actually lets you adjust brightness on a desaturated image based on the original colors. For instance, say I have a red brick with blue mortar. When I fire off the adjuster, I can darken the blues on one channel and the brighten reds on the other. Mix that with selective color, which I use to adjust my whites, neutrals, and blacks on the B&W image, and you can get some great heightmaps with only a minimal amount of work.
If GIMP has something like that, then use it. If not, then I suggest you use all your time and energy tracking down a plugin. It's that useful.
