this is a really neat gif but also a great example of the property that arises in 3D modeling… ”deminishing returns”.
every model is made up of polygons/triangles, which are a series of 2D/flat planes that can only bend at the edges where they meet. Every polygon(which is made up at minimum of 2 triangles) costs CPU power, and in things like video games, that CPU power is critical to preserve.
In the guy farthest right, all you see are these triangles, and he’s almost totally obscured. The 2nd one is a little better… the third is better still, but by #5, you’ve got the shape of a human pretty well done, and by the last two on the left you can barely see the difference. After a certain point, doubling the count of polygons or triangles that are allowed to make up this model doesn’t really matter… it won’t ever look any better, and if it does it will be barely noticeable.
here’s another neat example:
see? the difference is that 6k tris and 60k tris, they’re not really… that different. But the CPU power required for them is HUGELY different. So, obviously, you know the choice to be made here.
How do 3D characters not look so ugly and blocky? we can’t have it look ugly, right, we want a nice smooth surface on our characters. Something called subdivision comes into play. Subdividing, by the literal definition would mean that you take those tris and cut them in half, again and again, making more tris, until it resembles a smooth surface, but that’s insane and would take massive CPU power. So some guys back in the 80s on really dinky-ass computers figured out how to make it look like you’ve subdivided the shit out of a model so it looks like 60k tris, but isn’t even a quarter that, two guys you may have heard of: Ed Catmull and Jim Clark. Yes, that’s right, Ed Catmull, as in, the former CEO and co-founder of Pixar!! Back then, Pixar was just the name of the powerhouse used for Lucasfilm movies. They made tech demos, like the famous Luxo Jr, to showcase this Catmull-Clark subdivision.
In a nutshell, this is done by calculating the angles the sides of these tris meet, and making a curve in space to describe what the average would be if it was describing a surface. You can see the difference here:
anyway, that’s how 3D models look the way they look, and how they’re modeled. Using this, and planning for it while you model, can make your models look amazing and awesome without much computing power… people have experimented over the years to create the best “topology”, or, how these tris are arranged to provide the best-looking final smooth result when the subdivision is applied to it. And people are still learning how, and finding out new techniques, and getting better!
