First up have a gander at this beautiful tech demo that uses DirectX 11’s rasterization and post-processing engine to produce some truly spectacular looking images:
If you want to have a play with it yourself you can download the demo from the creator here. The recommended specs for it aren’t exactly mainstream though so it might not run particularly well on hardware that’s a couple years old now. Whilst I haven’t had a go of it myself yet there are a couple things that I picked up on in the video that I think are interesting, mostly because they give us some insight into how we perceive things as being realistic or not.
For starters you’ll probably notice the large amount of blurring that’s present in the video with the only portion of the image that’s always clear being somewhere in the middle of the screen. In photography terms this is referred to as the depth of field of an image or put more simply the area of the image that’s in sharp focus. For the tech demo the depth of field is incredibly shallow most of the time which, whilst not being done in order to make rendering faster (it in fact makes it slower as the whole scene has to be rendered then blurred), it does have the effect of making the scene look at lot more realistic than it would otherwise.
Now I’m not completely sure why depth of field works so well for making generated images appear realistic but the research I’ve dredged up seems to indicate that it has some to do with depth of field being used by our brains as a visual cue. Since most generated images are viewed on a 2 dimensional plane there’s no real depth to any of the images (does a picture of a hallway look like it’s continuing on through the monitor?) but replicating the depth of field seems to trick our brains into thinking it’s more realistic than it is. I’d bet that the demo wouldn’t be looking anywhere near as good if the depth of field wasn’t present and I’ll have to give it a look over to see if there’s an option to turn it off for comparison.
There’s also some other sneaky tricks being used in there like the use of rigid bodies for all of the things being rendered on screen. Whilst the physics appears to be very realistic it’s probably the simplest kind of interaction to model given that a hard object on a hard surface is about as close to ideal as it can come. As far as I can tell there’s either only a single or small number of lighting sources present in the scene which would make the whole thing a lot faster to run.
All this being said however it doesn’t detract from just how good this tech demo looks. There may be a ton of tricks used to get this looking the way it does but that’s what computer graphics are all about: finding those tricks and then using them to our advantage. Whilst you probably won’t be seeing many games based around this kind of tech for a while (the system specs alone are prohibitive) it does show what current generation computers are capable of producing and should excite you about the future possibilities.