(Disclaimer: I have no current plans to make a CG model of the Enterprise-D. I'm just curious about modeling techniques.) How would a CG modeler tackle the engineering hull of the Enterprise-D? I've been reading a little about techniques for modeling organic surfaces -- like NURBS and subdivision modeling -- but these techniques seem more "free-form." Trying to recreate a pre-established shape, with a precise contour, would be difficult with these techniques, wouldn't it? On the other hand, I've not tried these techniques, so I wouldn't know from experience. Opinions? How would one make Enterprise-D's engineering hull, with its smooth upper and lower surfaces, a sharp-edged "undercut," a thickened "ridge," and precise contours on all of these surfaces? Thanks in advance!
For more detail on what... those... are, and how to use them, check out Kier Darby's tutorial on how to model a White Star from Babylon 5.
How do splines differ from NURBS? My undestanding is that NURBS is a special (as in N-Uniform Rational B special) version of spline modeling. Do you have any recommended tutorial sites? The reason I'm asking is that, while I don't plan to model the Enterprise-D anytime soon, I have an original design which exists only as sketches, and it has many organic surfaces not unlike those of the Enterprise-D's engineering hull. Thanks again, and although I've said it before, your TOS hangar model is very cool.
Absolutely correct. NURBS stands for "non-uniform rational bezier splines." A "NURBS" surface is defined by, typically four (sometimes three) edges, each of which shares a vertex with one of the others. It then defines tangency conditions along those edges. Think of it like this. You have a four-sided shape consisting of bendable wire. You define the vertex locations, then the shape of the edge (the spline) that goes between those vertexs, for each edge. You then stretch a piece of rubber across the wire "frame." And finally, you adjust the starting angle of each piece of the rubber sheet at each piece of wire. It's how most complicated organic-looking work is done... at least initially. Now, for the 1701D's secondary hull, think it through. You'd need to lay out a "wire structure" of curves in space. Think of it like laying down the structure of the hull... each "wire" is a major framing element. If you can get the original construction cross-sections, you could lay these down, parallel to each other, at the appropriate locations. You'd then create NURBS surfaces over this "framework" and tweak the shapes at the appropriate places between "frame sections" to get the shape just right. You'd do this while mirroring your "tweaks" so that the shape tweaks happened, in real-time, on both sides at the same time. It's not hard, really... at least not in theory. The trick is that you really have to do your research up-front... either that or just be prepared to tweak, turn, tweak again, turn, tweak again... making the shape you have match the on-screen version by positioning it to match the shots you're working from and just "eyeballing" it. Not "HARD" exactly... but very time-consuming if you really want to get it just right.
Personally, I prefer Subdivision Surfaces. That's how I built about 75% of my Grandeur model. Sub-D's are not as precise as NURBS or what is usually called "spline cage modeling," but I find them to be more flexible and easier to control. If you have a template to work from and you're careful, you can get close enough that no one would ever notice the lack of precision.
One nice feature in Maya (and presumably in other packages, though I've never used it elsewhere) is the ability to convert a NURBS element to a sub-D element. So you can make the shape using the (much more mathematically-complex) NURBS approach, then convert it to a sub-D shape and keep 99% of the appearance while giving yourself a much easier simpler, NEARLY as visually-nice, and much less computationally-complex element. I don't use Sub-D's very much (and I never use them in my main modeling software which doesn't even have 'em). But I've done the "learning exercises" in Maya for sub-Ds and Vektor's definitely right, they're pretty damned powerful and can give quite nice results. Vektor, maybe you can give a brief primer on sub-Ds? My version is this... with sub-D, you model a polygon cage and then create "blobby shapes" that approximate the shape of the cage while being smooth (as opposed to true polygons which are, obviously, made up of a network of flat triangles and/or rectangles, and of absolutely no truly curved surfaces whatsoever). If you use 'em a lot, you can probably do a much better job as explaining them however.
The concept behind Sub-Ds is pretty simple actually. You create a low-poly "base mesh" in the approximate shape you want and then apply the subdivison, each iteration of which doubles the number of polygons, smoothing them out in the process. The catch is that you have to be very careful about how you build your base mesh and you have to have a good understanding of how the smoothing process works. For example, it is generally reccommended that you keep all of the polygons in your base mesh four-sided, which is often a total pain in the butt and requires a good degree of forethought. You don't have to do this, but the subdivided mesh is much more prone to pinches, distortions and smoothing errors if you don't. Another idiosyncracy of Sub-Ds has to do with sharp edges. By their very nature, Sub-Ds tend to round off sharp edges, unless you place two or three or more polygon edges relatively close together. These are usually called "control edges." In most cases, at least some rounding of edges is desirable because it gets away from those tell-tale, perfectly sharp edges that computer generated graphics tend to have. On the other hand, it tends to produce an inefficiently high number of polygons and getting a filleted edge of a precise radius is nearly impossible. It's possible to control which edges are smoothed and which ones aren't using the same smoothing groups that control rendered smooth shading and I often leave certain edges unsmoothed so I can go back and do it manually later, mostly because I can do it more efficiently by hand in terms of the number of polygons. Which brings up another aspect of Sub-Ds: You have to know when to use them and when not to, or at least how far to use them before switching to something else. A perfect example is the original TOS Enterprise. About 90% of that ship is far easier to build using standard modeling methods like lathes and extrusions. The Enterprise-D, on the other hand, is a prime candidate for Sub-Ds, especially parts like the secondary hull, because of all the irregular curves and organic contours. Regardless, though, I would never attempt to include things like the windows as part of the Sub-D object because building them into the base mesh while maintaining four-sided polys would be ridiculously tedious and the result would have an insanely huge poly count. There is a point where it would be far better to "freeze" the subdivided base mesh and then add the remaining details like windows using standard polygon editing methods. That's why my USS Grandeur still has hardly any windows, because those kinds of details need to wait until after the Sub-Ds are frozen, and I don't want to lose that flexibility just yet. Once you freeze a Sub-D and make additional changes to it, you can't go back without losing ALL of the additional changes. That should suffice as a primer on the subject of Sub-Ds, the entire scope of which can--and has--filled entire books.
The thing I don't like about subdivision modeling is that (at least in LightWave) the organic surface always winds up being slightly larger or smaller in volume than the original simple mesh. Spline modeling may be more difficult for some people (I actually find it easier than subD modeling myself), but at least the end result will conform to the exact spline outlines you've created. Having said that, I readily admit that subD modeling is the more appropriate choice in many situations, especially for organic modeling. (oh, and thank you FalTorPan )
True enough, but I’ve only experienced this problem with objects that I probably shouldn’t have been building with Sub-D’s in the first place. For example, I originally built the primary hull of my Grandeur model using Sub-D’s because it’s not just an oval, it’s more of an egg shape with the central axis off-center toward the front, and I thought it would be easier to do that way. Unfortunately, I ran into the very “shrinkage” problem you refer to, not to mention an excess of unnecessary polygons. I wound up going back and creating it with a lathe modifier, followed by a complex series of tapers and free-form deform modifiers. The secondary hull, on the other hand, is all done with Sub-D’s, and to the extent that I needed to match certain predetermined contours, I was able to simply stretch the base mesh until the Sub-D lined up acceptably. Another good example of Sub-D vs. Spline Cage techniques would be the B/C deck module of my TOS Enterprise model. The first time I tried to build it, I started with a simple lathed object and then applied a scale/taper modifier combination to the back half to get that distinctive teardrop shape. Unfortunately, the outline of the actual part is significantly more complex than that. Next I tried doing the whole thing with a spline cage, but again, there were some very subtle contours that I just couldn’t match. Finally, I lathed the front half and built the back half with a Sub-D, carefully adjusting the base mesh and the segments of the lathe so the two halves would match up as closely as possible. It’s all about picking the right tool for the job.
Your question is a bit like asking how to perform brain surgery. First you go to medical school. If you want to understand how to model something you have sketched, you are going to have to get a modeling program and learn how to use it to answer that question. As Tarter once said, "To know is to do". This is especially true because the approach you would take will depend not only on your skill but also which modeling program you are using and it's capabilities. Your choice of program will be defined by your budget and end goal for the model. You can use any of several different appraches to model: a) box modeling b) spline cage modeling c) nurbs modeling d) sculpture (such as ZBrush) In reality you will use all of these on a project if you are skilled/knowledgable enough and they make sense for a given "part". There are many good books on 3D modeling and you can find some good references on 3D modeling websites. Good luck.
Welcome aboard Waugh, you should show your Enterprise off. FalTorPan, are you planing on buying a program and begining the modeling process. Do you know which way you are leaning, or is this a generic question?
I'm a part-time hobbyist when it comes to 3D modeling. I own a very old version of 3D Studio MAX. I've played a bit with Blender, but I'm considering purchasing Lightwave. Most of the models I've built have been relatively simple, mesh-based models. I haven't worked much with spline-based modeling, subdivision surface modeling or NURBS modeling, which is a main reason why I started this thread.
Well, to help you a bit more... first bezier curves are merely one type of the more global category of curves called NURBS. From your perspective all you need to know is that NURBS are primarily used to model more organic (ie not made-made like) objects such as plants. However, they are often used by those skilled with them for other things for two reasons. First, the curvature is smoother (c2 vs c1) than a bezier which can be important depending again on the application (both what you are modeling and what you modeling in). The reason is that such differences will be hard to notice on certain forms and obvious on others. Basically, these are for advanced uses to get curves as smooth as a babys butt with smooth derivatives. Again, do not be mis-led about modeling. It is not you are a box modeler, he is a spline modeler, and she is a NURBS modeler - you are a modeler, and each of the above is a useful modeling approach to an individual part you need to model. One is more appropriate for one part than another and allows you to create a part more easily (ie faster). For example, imagine that you have to create the Enterprise TOS... You might do as I have done, and create the saucer with a lathed spline (line), use a spline cage to model the main hull, and use more of a box modeling approach to modeling the nacelles. You might however choose to use NURBS to model the hull and then convert it to polys later. In short, those are techniques to achieve your goal, and your job as the artist is to pick the one that allows you to model it best. Good luck.
WRONG. That is a total misstatement. NURBS stands for "non-uniform rational bezier spline." In other words, NURBS surfaces are built upon a specific type of spline... and a specific subset of the bezier spline category... to be precise, the subset of "non-uniform, rational" ones. You're confusing the child and the parent there. Otherwise, though, your post is quite good. Just needed to clear up that little bit.