3D vs 2D People: the Showdown!
February 14th, 2012 by ianOf all the objects in a rendering, people are the hardest to make convincing. That’s the result of thousands of years of evolution that has selected for our ability to recognize other humans, really, really well. Probably to keep them off our turf, originally, but that’s just speculation on my part. I’m not an anthropologist. Or a scientist of any sort. But anyway, in a nutshell, it turns out we’re really good at spotting when someone is a bit…off; much more so than we are with buildings, in fact.
Now, research has shown that where virtual people are concerned, as we traverse the spectrum from cartoon-like toward realism, our response moves from delight to revulsion, and back to delight as we behold real people. In other words, we have an inbuilt, negative emotional reaction to depictions of humans that look almost, but not quite, human.
Obviously, this is important where populating your rendering is concerned. The last thing you want is for people to be repelled by your condo-citizens. It’s bad for sales. So it goes without saying that we put a lot of thought into when we use 2D photos of real people and when we insert 3D, mathematically generated model-people.
The former of these are just as they sound: they’re regular photographs of living, breathing people that we insert in the last stages of the rendering. That is, they’re placed using Photoshop after the scene has been rendered. They look photorealistic, because, well, they’re photographs.
Three-dimensional people, meanwhile, use geometry and texture maps just like the buildings in the rendering around them (click here to see their component parts). If we rendered the same scene from a different angle we could show their backside—without making any additional adjustments! In short, they live in the scene with the buildings as opposed to be being placed as an additional layer in the finished image.
That last part sounds like a huge advantage, right? You might ask why bother painstakingly placing 2D photographs of people in the finished image when we’d have to do it all over again if we decide to re-render the scene at a different viewing angle? Not to mention the fact that matching the lighting conditions under which your 2D person was photographed to the finished scene can be tedious and difficult. Three-dimensional people are lit mathematically like the buildings around them and require no additional tweaking in this respect. They will automatically be lit the same and cast the same shadows as all other objects in the shared environment.
The problem is that 3D people take a really, really, really long time to model convincingly enough so that they get to the point where we don’t get “weirded out” by them. Long enough that, practically speaking, there’s no budget large enough to achieve populating your rendering with completely photo-realistic 3D people. Hair, for example, takes an astonishing amount of work to convince even the least observant among us that the person they’re looking at isn’t “fake”: rendering each individual strand takes crazy amounts of CPU power and time. So, if you’re ever managing a project with no deadline and no budget, this could be the way to go.
Mostly, we use 3D people in animations, since it would be close to impossible to do a fly through with moving 2D images: you would need a different photograph of every person in the scene for every frame! When we do use 3D citizens in these cases, though, we usually “ghost” them so that all the viewer sees are moving silhouettes. This gives us the ability to populate your animation while avoiding problems associated with showcasing humans that people could spot as fakes and have that almost subconscious negative reaction to we talked about above.
So there you have it. Unless you’re looking for an interesting Valentine’s date, being 3-dimensional isn’t always a plus.






