Nohiki, forgive the double post, but seeing what is here, I felt that it was a good idea to post a simple tutorial on one practical use of subdivisions. This is a very simple example, but it should demonstrate that subdivisions are somewhat labor-intensive when it comes to optimizing their usage in a model. Responsible modeling is good modeling...
That said, before I start, let me make some things known to you. I'm using AC3D, not Wings. However, I will speak in proper, general modeling terms. This should illustrate what's going on well enough that you ought to be able to see the process involved.
Now, let's start. We shall consider a cube, which everyone loves. Whether or not it is able to accompany you throughout your venture is entirely up to you - but enough of the lame game puns:
Let's say we want to subdivide this thing... unfortunately, doing so at this point in time would probably result in a sphere or an otherwise very-round-thing, and not a rounded cube. Thus, we must must do a simple surface (or face) division/slicing operation on the cube. In an ideal world, you'll see this:
However, we do not need all of those divided surfaces on the cube. They might not hurt the subdivision operation in their current configuration, either.
But, let's reduce the surface count to the bare minimums for what we want regardless - if nothing else, it's good practice. In doing this, try to keep in mind that a subdivision tries to evenly smooth a shape while retaining some degree of the base geometry. Thus, we need to give the model to be subdivided all the points it needs to keep at some level. Also, keep in mind that when subdividing a model, sometimes less is more. In other instances, you may need a lot of base geometry to get the desired result.
Now, apply subdivisions to the model until it's smooth:
Alright, it's nice and smooth, and it's just the shape we want. However, if we were to commit the subdivided model to proper geometry right now, we'd be producing a model which would be nothing but irresponsible for most applications - just look at the wireframe!
That said, we need to find a balance between the details we want to see and the details we can afford to see without straining our systems. That said, before commiting the subdivision (this might be AC3D jargon - I'm not entirely sure how Wings deals with subdivisions; however, all I'm stating is that although AC3D might be taking all of those surfaces/vertices on the wireframe into account right now, but model can still be reversed from its current subdivision level back to its base model format we had earlier), we ought to find a means or ommitting as many polygons and vertices as possible while retaining the shape. In the case of this example, we can easily slice this cube into eighths and then take one corner of the shape, reduce its complexity, and then mirror its geometry such that our end model will be a complete, reduced complexity, rounded cube. Here's the corner of the given cube, again, before committing the subdivision:
Now, commit the subdivision...
Just in case there's still some confusion, when I
commit the subdivision, that geometry now becomes permanent. The advantage of doing this at this point in time for this example soon becomes aparent - because a corner of the cube is identical on all three of its planes (X, Y, and Z), we carefully select identical numbers of surfces to combine into one. In my case, I select 100 quads (square/rectangular polygons) on each side, combine them into one surface, and then delete the unused vertices on the outer edges of the surfaces. Here is the result:
Now, if we apply this technique to the rest of the corner, combining sections of equal magnitude into singular, properly proportioned surfaces, we can attain the following result. Note how we are careful to maintain a surface of only quads, and how we maintain proper edge loops across the surface. If there's any questions on those terminologies or their importance, feel free to ask.
To finish this project, we mirror the corner along one of its interior sides by one of the axes, such as the x-axis. We then combine adjacent surfaces as well as deleting any vertices which are no longer necessary. We then carry on this process through the remaining axes until we have an optimized model. Here is the final result:
That's a simple example; there might be simpler means of doing the model as shown above, but the point of the exercise was to demonstrate how you can make realistic use of subdivisions. If we compare the final result to that of the initial subdivision, we might see (if we're really picky) a minor loss of detail, but in general, it's so small that no one should notice. More importantly, this new model is of a very reasonable poly count - the sbdivision model by itself was
wayyy to complex for intelligible use in most applications. And of course, if you have any questions, I'll answer them as best I can.
