In order to generate moving clouds, I have a giant square and I texture map a bunch of clouds onto it. I then increment the texture coordinates at each rendering pass. Is this a horribly wrong method of doing this ? It seems to be rather costly in terms of performance ? Any ways to make it better ? Should I use a different method like make small squares for each cloud and move them instead of the texture ? Nitzan ------------------------- www.geocities.com/nitzanw www.scorchedearth3d.net -------------------------

"In order to generate moving clouds, I have a giant square and I texture map a bunch of clouds onto it. I then increment the texture coordinates at each rendering pass."


Thats the way i do it. Either that or you could use the Texture matrix. The texture matrix might be easier if your using a vertex array for the clouds rendering primitive (be it dome/sphere/dodecahedron), but thats something i havent tried yet.


-=[ Megahertz ]=-

The texture matrix is very easy to use. With it you don't have anymore to write things like glTexCoord2f(coord_s + offset_s, coord_t + offset_t).

Moving textures has been discussed recently at the glTexGen Question thread.

About the one big quad or little quads, I think that :
- if the sky is very clear (2 or 3 little clouds in the air) you should use different quads for each cloud.
- if the sky is very dark (clouds cover almost all the sky) you should use one big quad.

That's from a performance point of view, not from a realism point of view (even though both methods are "reallistically" very close).

[edited by - vincoof on March 20, 2002 2:28:09 AM]