I used a couple different programs to achieve this effect.
Not that I think small divide is great as a desert, that kinda isn't the point. I'm talking about the texturing itself. I'm thinking with a little more detail here and there it just might work out.
First off, I like that you're using the same subject to study technique. Can I have the heightmap? I'll give it a go if I can dig up some time after I come back from vacation.
That main cracked-sand texture is tiling too much, imo- it needs some work on the main tiling texture to make it less obvious. I know that's hard, at some point you just have to do a lot of airbrush and pattern-stamp.
The mountains are interesting, but they don't feel natural to me. It needs subtleties added that need to be done by hand.
Doing a big rocky surface and having it feel right has got to be a real pain- I haven't tried it, and I don't know if I'd even do it half that well, so that's pretty much all I can say. It just doesn't feel correct yet.
Hmm, I like the basic theme, but have to agree that the flat desert texture is not varied enough, try adding some perlin noise to the alpha channel upon blending it with the other texture.
Also, I have found a very interesting way of greatly reducing the tiling of a texture:
Make a fast fourier transform of the image, and remove the low frequency components (these are the ones that cause the visual tiling effect), but dont remove the center dot (as that is what gives the base color), if its tiling strongly in only one direction, you only have to remove the frequency components of that direction.
Inverse FFT the image to get back the one with much less tiling. The results hardly modify the image, but the greaty reduce ugly tiling.
Argh wrote:First off, I like that you're using the same subject to study technique. Can I have the heightmap? I'll give it a go if I can dig up some time after I come back from vacation.
Uploaded the heightmap exporter widget to JJ.
Also, an example of FFT tiling correction:
Original:
Corrected:
The difference might not be noticable at this scale, but if you view the images after each other, the results are clear, the low frequency components are gone.
If you dont have FFT tools, post me the texture, and Ill have a go at fixing it.
Yar same tex, I've only used it twice I think, parched mesa and whakamatunga riri.
The obvious tiling is obvious. I was kinda more interested in whether the way it was textured could be considered valid or even viable...
Reason being, I was able to texture this stuff in 5 minutes using a combo of l3dt and cs3. Which asks the question, is it a valid way of doing things (assuming your tiling isn't omgwtfobvious?)?
It's kind of a rhetorical question, but one i would like an answer to in terms of opinions.
I think that if you got that far in 5 minutes, great- automation is cool! Too bad it requires CS3, I only have 7, and I can't afford to upgrade it any time soon
Oh, and I'd like to hear more about the FFT stuff. Where can we get tools to do that with? I'm accustomed to just doing this stuff the hard way, so anything that could speed up that part, which I personally find a total drag, could be useful.
I would love it if you added some mountains encircling the area(make the map a bit larger then add mountains. The cut off is harsh for screenshots. Other then that I wish that this and your other remake of this map were included with spring they are very nice.
Here's the deal. L3dt renders things by use of various maps. The 2 important ones here are the attributes map and the terrain normals map.
With the attributes map you get a basic color of what texture will get laid where. This is important because you can just do color range selections in photoshop and apply patterns on different layers. It also means that because you are using blending options to display the pattern, you can scale your map to whatever size you want. Like small divide here I could easily make into a 28x28 simply by resizing it without losing detail.
The normals are important because you can bake them as an overlay, and you can even tell l3dt to do the light based bumpmapping directly on the terrain normals (in this exercise I didn't do that, so I'm curious to see what will happen when I try it). This way you get the natural variations which breaks up the terrain in a nice way.
Small divide's heightmap isn't anything to get excited about honestly, but in this example it was a good familiar reference. As a 1 man dev team (except all the lua that lurker has done for me), I have a gold of 20 maps just for evolution that will be included with the game. As a result, I have to barter time and cpu resources. This way I could prolly do a 20x20 map in a matter of minutes.
THink about it. You can import the attribute map at 512x512, layer the hell out of it, the resize and save as bmp, meaning that you can do almost all of the technical details with a small image and therefore quickly. Use features for metal spots (if needed) like what I used on riverglade, and all of a sudden the hardest and most time consuming processes of mapmaking just became easy as all get out.
Now, does this replace texture generation? In my mind no. However, using this method you could pump out maps with quality just as good (assuming you took the time to set up your textures correctly like behe was talking about) quality, and even save a bit on filesize. Hell even doing strata would be easy using an overlay texture.
However, the only reason this method works is because of l3dt. Without that you're essentially up shit creek, because you NEED the attributes maps. However you can render the attributes maps at a fairly low detail (2048x2048 should be enough).
This could be a saving grace for me, cause I have been frustrated about it for some time now.
I still think we need an sm3 sort of mapping thing, but until then, this might just save my bacon, because I simply do not have the time required for texture generated mapping.
Argh wrote:
Oh, and I'd like to hear more about the FFT stuff. Where can we get tools to do that with? I'm accustomed to just doing this stuff the hard way, so anything that could speed up that part, which I personally find a total drag, could be useful.
Personally, I think the best effect for flat terrain is obtained by making a set of interchangeable tiles and randomly selecting - then you get tile compression without a repeating pattern.
Before blurting out "LOLTILE" I wish you guys (the people that blurt out "LOLTILE") would actually do some research.
If it doesn't repeat, it doesn't compress. Capice?
Can't it? I thought that was the point - if you have a tile that is a repeat of another tile in the map - even one that isn't adjacent - then it would compress it.
Damn it, I'm a few hours into a tool to make tiled maps easy, for nothing?
Sorry to burst the bubble, but that's about the size of it. The compressor is really really bad. Also, if you make maps obviously tiling, then they pale in comparison to "Real" maps.
This test was an excellent example. It works, but it doesn't work well. The only way to really break it up would be to use a massive bumpmap over the entire map so that the terrain break up, but guess what? That breaks the compression as well.
Besides, almost all of us compile with compression off because it makes maps look terrible even on .1
It's not like the compiler automagically knows the size of repeating texture you use anyway, so it does a really crappy comparison, and fails miserably at it.