Sometimes you get the idea for something “new” be it real or otherwise from a problem you’ve run across. Other times it’s simply from doodling out random ideas that start as simple geometric shapes. This project is a marriage of both of those sources of invention. Unfortunately, until I get a patent, the more interesting part of the project will have to go unseen for a time. That said there is still some minor designing and 3D technique to geek about!
So as I said it started as a sketch. I’m a big believer in “doing” something is to understand, and sketching, no mater how poor you are at it, is a great aid in conceiving or refining ideas. I also said that the more interesting part of the project can not be shown yet… unfortunately that includes some proof of concept sketches. That said here is an early doodle of “ideas” for what it’s worth.
On the 3D side of things, this is not really the most remarkable mesh. True, it has some fairly proficient sub division surface modeling in areas, but to be honest, it’s nothing to toot ones horn over (or honk, beep, chirp…). I knew I wanted to keep the edges of and profile shapes of the mesh clean and simplified even though a real bag would be more “shlumpy” and have more folds, seeing as this was a concept piece. The interesting part is more in re-creating the fabric(s) for this piece. I chose two fabrics to work with: a ballistic fabric that is found in many bags now (such as timbuk2 models) and a less conventional, more for fun treated upholstery type wool blend. There is also a retro-flective material which can be seen here as the stripes on the front of the bag at full reflection incidence. Here are a few close ups of those fabrics:
The shader trees to these, while not mind blowing in complexity are some what “tricky” esp. the one for the ballistic fabric. Due to the anisotropic nature of the fabric this is a surface that can not just be “photo pasted” on with a scanned example of the fabric and a bump map, a practice which I despise anyway as it never replicates the true feel of a fabric and how it reacts to light, and while fast to setup, is highly limiting in the end. To re-create this fabric correctly, I referenced first the creation of multi-directional ansitropic highlights as driven by a UV directional raster map, found here on XSIblog. If you’re simply creating a metal, normally you can stop at a halfway simple tree with this technique, but you’ll need a bit more to get your fabrics right. Here is an example of my render tree in Softimage 2011:
A few of you might see that I have several normal maps that I’m adding in the tree, and might wonder why I’m doing that instead of rolling them all together. The reason for this is because some of the maps are effecting large areas, and need to be high resolution such as those for the folds/creases while the basic bump for the ballistic threads can be simplified and tiled. This saves memory and provides extra flexibility in this instance. (It should be noted that not all applications would require such complexity in their surfacing tree. Messiah for example, due to the highly flexible nature of it’s surfacing tree that provides mixing and math operations right on the input between more than one entry, allows this tree to be created with only a small handful of nodes and one grayscale/scalar map input driving the direction of the anisotropic effect.)
Here is a shot of the basic bag’s folds being roughed out and created in 3D-Coat using normal painting tools.
I chose to use direct normal mapping because I knew I wanted to hint at folds and creases, but not actually distort the geometry. I wanted to present a clean, concept look with little detraction from the fabric, function, and colors. The bag also needed to look sturdy, and not like the average book bag- so again, preserving profile edges was paramount. Besides, for details like stitching etc. there is often no need to create a highly dense voxel or polygonal mesh model for such small details to only end up baking out a low(er) resolution normal or displacement map.
It also had the bonus of, unlike distorting the actual geometry, being easy to edit out or tone down in PS if needed, and being much faster to render than displacement. Later I would blur and further blend areas in PS and exclude many of the folds in the front since even though they added more realism the ultimately were to detracting.
The texture for the treated wool bag is actually procedural do to the need for a very “random” look to almost all aspects of the fabric. I could have baked out several layers of maps and then re-applied them (saving some render time), but since this project was not to be animated, and I wanted as much detail as I could get I kept it all as a procedural tree.
Well there is the beauty and horror of nodes and the ability to re-use and sample things which is powerful, but sometimes messy, especially if things are not labeled under a convention of some kind. This is especially true of this tree more than the first, with so many things driving one another and branching so frequently. In reality, while the tree looks complex, only two of the nodes need minor alteration to control the fabrics color and this could also be more streamlined with the addition of several math and color correction nodes that act as inputs for a compound.
OK, enough geek speak for now. Hopefully I can get my patent through and show the more interesting part to the project soon- wish me luck!