|
- About |
|
Thinking
Outside the Box:
By
Paul "Villam"
Steed
Loft / Extrude When lofting or extruding something you're basically just taking the outline of a shape and lofting it along a path (i.e. an extrusion of that shape). This technique is very useful when a primitive won't fit the bill. Lofts can have as many segments or layers as you require.
Primitives Primitives are a term used for a group of polygons that can be created quickly by most modeling packages via a single command like 'create/box' or 'create/sphere'. Primitives are excellent for starting a model and can be likened to a lump of clay or putty. If primitives can't fit the need of a shape, then of course a loft or Boolean would be next.
Boolean You'd think that Boolean would be the coolest of tools when creating models since it alters the shape of an object by using the intersection, subtraction or union of another overlapping shape to subtract or add to the original object's geometry.
In reality this tool can be a pain in the rear since (at least in 3D Studio 4) it creates extraneous faces that are sometimes hard to find thus adding to your face count unnecessarily. Still it is very, very useful when it works like you want it to. Level of Detail (LOD) This is when only several triangles represent the object if it's no more than a few pixels on the screen, and the closer you get to the object, a more detailed model pops into view. This technique can be done manually, but if you go up to say 10 levels of detail you'd take a memory hit because 10 separate objects would still have to be stored and tracked instead of one. A technique called 'displacement mapping' or 'real-time deformation and tessellation' based on displacement map information can overcome these LOD limitations but has yet to be implemented on a large scale. We will be using LOD in Quake 3:Arena, but each LOD will be done manually by yours truly with the distance of their respective appearance dialed in appropriately. 'High to Low' vs. 'Low to Higher' These are two methods by which I model based on how I'll be using the mesh. The former method is usually reserved for meshes that will be in high-rez cinematics as well as in the game. Initially an unlimited face-count approach is taken and the model will end up being very detailed. This high-rez version can be used for advertisements, cinematics or whatever. It also serves as the template to model a low-rez version of the same model to go into the game. The latter technique involves creating a mesh without considering a high-rez version and keeping the face-count relatively low from the beginning and optimizing on the fly. Sometimes, though I'll make a high-rez version of a character or object just to have a starting point to texture. Usually Adrian is our texture guy so I don't overly worry about the texturing process. Here's a comparison of a hi-rez Gunner I built from Quake 2 to the actual in-game Gunner…
As you can see the low-rez version wouldn't hold up very well on the cover of a magazine despite his charming demeanor. Here's a better comparison of their relative difference in complexity…
|
