Um.... what do you expect really? The wider view the more stuff need to be included in the same viewport. Of course the stuff will look smaller ("moving away from the camera"). How do you expect to place the much more data in the same 2D space? Could you post a simple drawing what do you expect the result to be?

Or you want the end result to be like this?

If so, I'm not sure how you can achieve that. Maybe you could use a lot of frustums to render it, or maybe some shader trick would suffice. Look into "sphere mapping" or "sphere map rendering" or something like that.

If you're not aware, 3D is mapped to a 2D-plane (at a fixed distance), the greater the FOV, the bigger this 2D plane becomes (at 180 degrees it would be infinitely large)... additionally, as you sample further from the center, the angle increments in 3D space becomes increasingly smaller... which causes it too look very stretched at the edges.

http://www.cs.ucl.ac...es/image006.jpg

However, you should remember that a 3D world that is rendered to a 2D-plane should optimally be viewed from a certain distance from the screen (depending on the FOV). Looking at a rendered image from any other distance will cause varying amounts of perceived "distortion" or "weirdness". Especially if you look from further away with a very high FOV, which gives the traditional "tunnel vision" look.

As for rendering the "same image" with a different FOV, I see nothing wrong with the 2nd image, you can clearly see the 1st image in the middle of the 2nd... so I'm not quite sure what you expect to see? Perhaps you intention is to just increase the "horizontal FOV", so to speak. If so, lower the height of output. Anyway, FOV is the same thing as zooming, or looking through a binocular if you decrease the FOV, or if you increase the FOV it's the same thing as looking through the binoculars from the wrong side (everything becomes very tiny).

Perhaps what you are looking for isometric view? That is, removing the perspective in favor of keeping a consistent scale all across the output image, meaning, you can no longer determine the depth to an object, but there will be no visible distortion from huge FOVs. Or perhaps you might want something in-between, keeping a very low FOV, but instead moving the camera a lot further back (still provides a sense of depth, but far less distortion).



As with the laying out the world map, there is exists no "perfect solution" for mapping a 3D object to a 2D image, there are a bunch of different solutions for different purposes.

Like the poster above, a better description of what you expect would really help.