I've asked this before, but I have a conceptual solution so I was wondering if someone with a better knowledge of graphics libraries and wacky new features of G-cards could point me in the right direction. I have a terrain in which there a lots of overhanging cliffs - basically it does not approximate to a heightmap - but it has many of the same rendering issues: it is a lot of triangles, I am repeating a tileable texture over it. Now, if I drop textures onto it from above I get those nasty vertical stretch marks, and I want rid of them. Best I have so far: multi-texturing with 3 textures, one along each axis. Pros: easy. cons: stretch marks still visible, just less obvious. What I'd like is a way to make the alpha of the textures depend on the angle of the terrain at which they hit - I would apply one texture all over, on the X/Y plane, and then apply the next on the X/Z plane but with an alpha of 1.0 where the surface is parallel to X/Z and an alpha of 0.0 where it is perpendicular. Then I would repeat for Y/Z with a third texture. However, oldstyle (which is all I know) that involves calculating a whole heap of material colours for each vertex. All I know about shaders is that they sound quite pretty... Im happy to learn something new, but could someone suggest the alternative approaches I might take to get the effect I'm after? I don't want to learn GL slang only to find that a) my graphics card doesnt support it and b) I could have done it in 3 lines of code with pixel shaders. I hope people understand what I'm asking :) Thanks

yeah i suppose that would be a piece of ceak for a pixel shader.

doing it with vert. colors would require multiple rendering passes, no? but it should be just as possible. just make sure the parts sum to one, which you can simply do by multiplying each component of the normal by itself. should give the desired effect i think.

Indeed, a pixel shader would be the easiest way. I think it could be possible to use the dot3 extension (per-pixel lighting) to blend between three textures without shaders in one pass but it's limited to black and white... So it'd be just a detail texture and you'd have to supply the color per vertex or something. Most likely you'd still need at least two passes.

Oh, I know OpenGL, and Im sticking with it since I like to be cross platform. Sounds like the general consensus is pixel shaders. Google is almost overcome with enthusiasm in answering a search for tutorials - looks like CG is thw way forwards for a C++ programmer, unless OpenGL 2.0 suddenly becomes standard?