My first CMOD sprites models

[Cham]

Here's a nice one : the galactic halo. The model behaves perfectly on my system (except that the blobs shouldn't be scale invariant. They should change their apparent size with distance). Edit the toggle script file and the model's color as you wish. Need feedback to know how the model is behaving on other systems.

halo1.jpg

halo2.jpg

Galactic_halo.zip

[t00fri]

Martin,

after coding the Celestia galaxies and globular clusters in terms of sprites, I have plenty of insight about what really matters in this context.

I know little about and I am hardly interested in sprite renderings via the add-on path. Still things cannot be very different.

I think the add-on sprite scheme lacks TWO essential aspects to allow for really good-looking structures: In EVERY model point you must be able to set the SIZE of the attached sprite texture as well as it's brightness! You seem to be working with a single size of your sprite texture, which -- no wonder-- then leads to such dull representations:



Fridger

[Cham]

In EVERY model point you must be able to set the SIZE of the attached sprite texture as well as it's brightness! You seem to be working with a single size of your sprite texture

Fridger,

currently, AFAIK, Celestia isn't able to draw the sprites with a well defined size, dependant on distance (unless there's some option I don't know yet). I already asked Chris for this, recently. He didn't responded yet.

I already tried various sprite sizes in the CMOD files, and yet I'm unable to notice any difference in Celestia. Celestia's sprite code is simply "archaic" and "primitive", in the case of CMOD objects.

[t00fri]

In EVERY model point you must be able to set the SIZE of the attached sprite texture as well as it's brightness! You seem to be working with a single size of your sprite texture

Fridger,

currently, AFAIK, Celestia isn't able to draw the sprites with a well defined size, dependant on distance (unless there's some option I don't know yet). I already asked Chris for this, recently. He didn't responded yet.

I already tried various sprite sizes in the CMOD files, and yet I'm unable to notice any difference in Celestia. Celestia's sprite code is simply "archaic" and "primitive", in the case of CMOD objects.

I know, since I looked into the code before my previous post.

But what I wanted to express is that in this situation it will be hard to achieve good results. It's not only the dependence on distance that is missing. You need a ~ statistical interference of sprites of different sizes and brightness for achieving a natural looking result. Also you would need to write some .celx code to get the right modulations of sizes and brightness. This might well become VERY slow. One has to do these things on the C++ code level, I am afraid.

Fridger

[chris]

CMOD point sprites do in fact have a well-defined size that has the same units as vertex positions. I'm wondering if there's another driver bug at work here. The point size calculation is actually performed in a GLSL shader, which is perhaps why Fridger didn't find it. In shadermanager.cpp, there's the following function to calculate the final sprite size in pixels:

Code: Select all

string
PointSizeCalculation()
{
    return string("gl_PointSize = pointScale * pointSize / length(vec3(gl_ModelViewMatrix * gl_Vertex));\n");
}


Here, pointSize is a value that's defined for each point in the model, and pointScale is an overall scale factor based derived from the DSO or solar system object's radius. As you'd expect, the sprite radius is scaled by the inverse of the distance (of the transformed vertex position from the eye position.)

Martin, your model looks significantly different on my system: although the sprites do shrink with distance, they are much too large. The Milky Way appears to be hidden in a nearly opaque blue-purple haze. I suspect that ATI's drivers are still messing up point sprites. In my work on particle systems, I've abandoned using hardware point sprites and opted for software generated sprites--they're slower, but they're not hampered by driver troubles.

--Chris

[ElChristou]

If you want feedback, better post the models, else...

There's a test addon, on the previous page!

Oups!