Hi,
I've had this idea for some time now, but since I have almost no experience in realtime 3D programming, I wanted to know if it seems possible.
The idea is to have a top-down view (cf. GTA1) and to compute which areas are not visible for the player. I thought that lights and shadows algorithms are probably well optimised so here's the idea :
-replace the player (the unit on the ground) with a light.
-render the image to a buffer
-threshold
(-apply some sort of blur)
-use this buffer as a mask for the final render of the scene.
Of course the first render could be done on a simplified scene.
This would be used at first for a small hide and seek game.