The author of chapter 6 says on page 201, "It is too early to tell what this current decade will be dominated by, but my prediction, especially for the next 10 years, will be procedural object and surface definition. We finally know how to render any model; now let's make the computer help us build models themselves."
Procedural modeling. Hmmm. Interesting. This book was published in 2003, so we have some time for this to happen, I guess. What do you think? I'm curious, since this book is well over my level of mathematical and programming capabilities, but I find these sorts of observations interesting.
Well, we've got procedural, landscape generation, (TG2 is but one of several each with their own distinctive feature set), theres a number of tree generation systems, (mostly variants of L-systems) and I've seen a few facial and human body modeling programs that use a genetic variation to make up different characters. The trouble with all these systems is twofold, firstly that the system has to be "taught" in some manner what results are acceptable and secondly that they all work on a trial and error approach. If you hit the random seed button enough times your bound to come up with something interesting. I'd like to see a system where I could "rough" our a basic landscape, tree, body shape etc telling the program that I want a mountain, valley, tree limb or body part in a certain place and the fractal/procedural detailing would take over from there. I think the latest iteration of World Machine allows you to map out mountain ridges (haven't updated in a while so I could be way off) and this would be a step in the right direction. Another option I would like to see is "fractal isolation" and blending where I could grab a feature from one procedurally generated landscape and use in in another. I think we're a way off fully procedural modeling but we're on the way.
Richard
I can't find the thread, but procedurals towns / cities where discussed in it and some links where provided. The images looked very very good.
You can find informations here:
http://citygen.net/
http://forums.introversion.co.uk/introversion/viewtopic.php?t=586
http://graphics.ethz.ch/Downloads/Publications/Papers/2001/p_Par01.pdf
For trees:
http://forums.introversion.co.uk/introversion/viewtopic.php?t=602&highlight=subversion
@Realuser - The cities were discussed in this book as well. It sounds fascinating.
@cyphyr - The real question I have is whether learning modeling applications will become unnecessary in the next 10 years. Modo, Lightwave, 3DMax, Silo...all a thing of the past. Or not?
Explicit modeling will probably always be necessary. It will be assisted by procedural methods, perhaps to greater and greater degrees, but procedurals will likely never get to the point where you can control their shape explicitly enough through indirect controls to get the result you want. Learning traditional poly-based modeling tools may be less useful, but the skills of hand modeling in general will definitely be useful. Remember that there are also many objects that do not lend themselves to modeling by procedural or "organic" (e.g. ZBrush) methods, so poly-based or some other form of "rigid" modeling will always be useful.
- Oshyan