Slashdot Mirror
POV-Ray Competition Winners
An anonymous reader noted that you can "See how far POV-Ray developers have pushed the limits of raytracing in the POVCOMP 2004 Raytracing Contest." Yes it's from 2004. It's still neat. And you try finding something interesting on a holiday monday ;)
Of course, no article on POV-Ray is complete without the obligatory link to the site of Monsieur Gilles Tran, surrealist and POV-artist extraordinaire...
Has he entered the competition? Haven't seen his name anywhere so far...
Tedious Bloggy Stuff - hooray?
...to make something photorealistic you need to create extreemly high-poly models, plus you need either humungous texture files or to write a dynamic shader. All that takes lots and lots of time.
The only thing that makes that office render not photorealistic is that a lot of the textures are too "perfect" for want of a better word. Look at the filing cabinet in the background, if this was a real office there would be lots of tiny dings and scratches. That kind of thing takes a lot of time to model.
I am NaN
I'd say a little of both.
I read a very interesting interview quite a while ago in (I think) a Wired magazine article. The topic of discussion was the creation of realistic 3D human models. One point, if I recall, was that you have a lot of leeway as you're moving toward a realisting image, but once you cross a certain line, the absence of the most seemingly benign details will give it away.
I think the same applies to modeling in general. Take the office image for example. The lighting is very good - if you look along the edges where the walls meet the ceiling, you'll see subtle light "spots". It's not that this is anything unique, but that they were rather well done. They're subtle - if they were missing, you might not notice at first, but I can pretty well bet that it would still register- not as something that would be readily identifiable, but something that's just "missing".
If you look at something in real life, and you set yourself to reproduce an exact replica, you're forced to deal with the collective imperfections that make the object what it is. Suffice it to say, straight edges are rarely perfectly straight, but 3D modeling makes it exceedly easy to produce them as such. The challenge is introducing just the right amount of imperfection.
Add in lighting - that's often something that will make or break an image. In fact, lighting is so important (imho), and getting it "right" takes a lot of time and tweaking. When you factor this into the length of time required for a good test render, you may find yourself settling for "not exactly what I want, but good enough."
So, it's a combination of things. Even if someone had a supercomputer at their disposal, I think you'll still see a lot of work that comes close, but just slightly misses the mark for one reason or another.