device.TextureState[0].AlphaOperation = TextureOperation.SelectArg1;
device.TextureState[0].AlphaArgument1 = TextureArgument.TextureColor;
device.TextureState[1].ColorOperation = TextureOperation.SelectArg1;
device.TextureState[1].ColorArgument1 = TextureArgument.TextureColor;
device.TextureState[1].AlphaOperation = TextureOperation.SelectArg1;
device.TextureState[1].AlphaArgument1 = TextureArgument.Current;
device.SetTexture(0, texAlphaGrass);
device.SetTexture(1, texGrass);
device.DrawIndexedPrimitives(PrimitiveType.TriangleList, 0, 0, numVertices, 0, numTriangles);
device.SetTexture(0, texAlphaDirt);
device.SetTexture(1, texDirt);
device.DrawIndexedPrimitives(PrimitiveType.TriangleList, 0, 0, numVertices, 0, numTriangles);
Single-pass Texture Splatting
Author
Maybe this is impossible and I should just do multi-pass...
I'm wanting to splat a grass texture as a base, followed by an alpha mapped dirt texture. When I do this:
it works! texAlphaGrass is just a white .dds...is there a way to not have to use this alphamap?
However, that's multi-pass, and what I want to do is use TextureStages 2 and 3 to set another alphamap and dirt texture, removing the need for another DIP call. When I try it, setting up the TextureStages the same as 0 and 1, I just get the base grass texture showing...my textures 2 and 3 don't seem to have any effect.
So, does anyone know if/how I can do this in a single pass?
Cheers, James
[Edited by - jameswren on July 9, 2005 4:52:00 AM]
As far as my research showed, you cannot do texture splatting with alpha maps in a single pass. Since the single-pass blends everything all at once...
Oh Indeed.
A shader will allow a single pass no problem, especially cuz you can compact alpha's into each color or a texture.
http://www.gamedev.net/columns/hardcore/splatting/
Read up on that article.
His explanations are said better in the text he refers to at the beginning.
His method is an upgrade tho since it use's a pixel shader. The PS code is provided, I am using it now since it is very small and effecient.
Remember tho, it only shows the most basic approach to texture splatting, it can be way more.
A shader will allow a single pass no problem, especially cuz you can compact alpha's into each color or a texture.
http://www.gamedev.net/columns/hardcore/splatting/
Read up on that article.
His explanations are said better in the text he refers to at the beginning.
His method is an upgrade tho since it use's a pixel shader. The PS code is provided, I am using it now since it is very small and effecient.
Remember tho, it only shows the most basic approach to texture splatting, it can be way more.
I have tried to do this as well. Unfortunately, there is no way to do base textuer + texture splatting using the fixed function pipeline in one pass.
I believe it would be possible with vertex alpha. Just change the texture stage state for the second texture to use the vertex alpha component rather than the texture alpha.
Author
Thanks everyone.
If I use vertex alpha, I guess that would give me the same alpha resolution as my heightmap, so would this be worse quality than using say a heightmap of 1024 and an alphamap of 4096? Although would that loss in quality be acceptable due to the gain in speed by doing just a single pass instead of 3 or 4?
If I use vertex alpha, I guess that would give me the same alpha resolution as my heightmap, so would this be worse quality than using say a heightmap of 1024 and an alphamap of 4096? Although would that loss in quality be acceptable due to the gain in speed by doing just a single pass instead of 3 or 4?
Well, you can see some snaps of my engine here and here, as well as in here. I seemed to have posted a lot of picture of it [wink]
Keep in mind that with vertex alpha you won't be able to do non-linear blends between textures, ie something like this
Actually, the blend from one texture to another does not need to be any more linear with vertices than with pixels. You could achieve the same result with a decent resolution mesh. Usually, the alpha map would need to be pretty huge in order to have a resolution that couldn't be handled by vertices. So unless the vertices are really spaced out (like my system - 32 inches apart), or the alpha map is mega huge, you can get just as pretty results with vertices.
I think the alpha map management is a pain. It also hogs memory. Every single layer needs it's own full-scale map, other than the base. So if you want cool spotting effects, your video memory requirements are going to shoot into the sky.
I've also thought about trying an alternative system. Where a precalculated (drawn in a paint program) alpha map per transition type could be used on different areas. For example, you could have a corner section, a straight section, a blurry section, a speckled grass section, a hard edged corner, etc. They could all be stored in the same alpha texture, but each quad would use it's own coordinates (secondary alpha coordinates, where the primary coordinates still tile across the landscape) to reference the mask it wants to use. It should allow you to use extremely high res alpha values for small parts of the terrain, since each quad can share the same texture coordinates. Allowing you to literally provide any shape or pattern you want, specific to any layer or texture, in extreme resolution. At the same time, non-faded quads could be rendered with tiny, totally opaque, single pixel coordinates to speed up the rendering. The most difficult part would be building an editor that used all of this flexibility to make an easy job out of building the terrain :)
I think the alpha map management is a pain. It also hogs memory. Every single layer needs it's own full-scale map, other than the base. So if you want cool spotting effects, your video memory requirements are going to shoot into the sky.
I've also thought about trying an alternative system. Where a precalculated (drawn in a paint program) alpha map per transition type could be used on different areas. For example, you could have a corner section, a straight section, a blurry section, a speckled grass section, a hard edged corner, etc. They could all be stored in the same alpha texture, but each quad would use it's own coordinates (secondary alpha coordinates, where the primary coordinates still tile across the landscape) to reference the mask it wants to use. It should allow you to use extremely high res alpha values for small parts of the terrain, since each quad can share the same texture coordinates. Allowing you to literally provide any shape or pattern you want, specific to any layer or texture, in extreme resolution. At the same time, non-faded quads could be rendered with tiny, totally opaque, single pixel coordinates to speed up the rendering. The most difficult part would be building an editor that used all of this flexibility to make an easy job out of building the terrain :)
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
