Hey, further to my experiments with dynamic dot3 mapping using a cubemap, nicely implemented by devil, I had a thought...
When changing the light source of a DOT3 map, using a cubemap, you have to rotate the cubemap so to speak. You place a camera inside a cube, textured with the normal colours required for a dot3. Then do 6 renders, to get the updated cubemap.
Now, the cubemap texture only needs to be very small. Say 2x12 for this effect to work.
Surely it must be possible, via some mathimatics, to recalculate the pixels of the cubemap, without having to texture a cube, and rotate it, then rerender the cubemap.
If anyone knows what i'm talking about, then do you think this is possible? It would be a hell of alot quicker!
|