Some 2D-3D visibility experiment
I had an idea for a simple 2d/3d visibility effect, and this is the result in Unity. Basically it applies the "soft shadows" effect from here to a 3d world.
Basically, what's happening is that there's two cameras: one 2d camera and one 3d camera. The 2d camera projects the walls into an "occlusion texture", and that texture is used to generate the 2d visibility texture. That visibility map is then rendered into a "cumulative visibility texture", which represents the black/white areas in the scene.
When it comes time to render the 3d objects, you pass the VP matrix of the orthographic camera to the shader and project every point in the orthographic cameras space. The UV coordinate of each vertex in the visibility textures is passed to the fragment shader. In the fragment shader, the interpolated value is used to figure out how to render by sampling the visibility textures.
Obviously, many optimizations and visual improvements can be made, the most obvious one being applying a circular blur to the visibility texture to give it a soft blur. There are also some annoying visual artifacts, mostly related to the visibility calculation algorithm (though those might go away as well with some blurring).
But this was just a weekend project for me, and I think it's kind-of a neat trick. With some work this could be used in a proper game.
Just a warning: I haven't one through and properly cleaned up the source and shader code, so it might be a bit messy. If you have any questions, let me know! My email is my github username @ gmail.com.
The code is MIT licensed, so you can pretty much do whatever with it, but if you do use it for something, I would really appreciate to hear about it!
