I have a rule: if a feature is going be used, it's going to be mapped.
I don't use solid colors; I don't use a single roughness value; I don't use a single layer opacity value; I don't use fixed angles.
Nothing in reality is perfect - no solid color is absolutely solid; no surface is absolutely uniformly rough. The greasy specular stuff exists on virtually every material I make, because it's on everything I touch. Sometimes it's more subtle than other times, but you simply can't overestimate the value of all these little imperfections. It goes without saying that my geometry is always similarly imperfect. I don't have perfect angles; I don't duplicate objects without heavily modifying the copies. I don't use snap aligns, I don't even model to ortho snaps. I do it by hand and by eye, and in my attempts to be perfect, the models are always gloriously imperfect, as reality is. Even the most rigidly machined product has visible inconsistencies and tolerances. It's the layers of ambient complexity, building up, which help sell the final image.
Another of my Rendering Commandments: Thou Shalt Not Use Thy Eye to Judge Materials. Use reference photos, only. Remember, in Maxwell, we are not recreating reality, we're recreating photographed reality, and a camera sees the world differently than our eyes do. I have this little toy car sitting next to me in my office. Looking at it with my eye, it looks different than the render does (not night and day, but still...). But take a picture of it, and it matches the render precisely. I have a number of HDRIs that I subject my materials to, which I have come to trust, and in the end, because Maxwell is recreating the look of low-dynamic-range capture devices (cameras), their images have to be our guides.
Changing a lens opening by "one stop" means either doubling or halving the amount of light that passes through the lens. By definition a .30 log neutral density filter will absorb 50% of the light passing through it (i.e. -1 stop)
If a cinematographer opens up "one stop" he is doubling the amount of light passing through his lens. To compensate he could put a .30 ND filter in front of his lens and be back where he started. If he's shooting under perfect conditions and opens up one stop and exposes a motion picture negative film such as 5247 then twice as much light will hit the film as should. To compensate for a one stop overexposure in the lab the gamma, or contrast, of the film must be considered.
The gamma of all contemporary motion picture color negative camera stocks is around 0.65. This means that for a one stop increase in exposure, the resulting density increases by .30 ND times the gamma of 0.65 which is .195. One laboratory printer point is 0.025 ND. Dividing .195 by .025 yields 7.8 printer points. Therefore to compensate for a one stop overexposure by a cinematographer one must increase the light in the printer by 7.8 printer points to make a print of the proper density.
If a cinematographer asks for the timing of a scene to be "one stop darker" he probably means, "make the scene look as if I stopped down one stop." So to accommodate the wishes of the cinematographer one should add 7.8 points of density.
If an optical house cameraman says that he shot the dupe neg "one stop overexposed" and asks the lab to compensate, the request is quite different: The optical cameraman is probably using 5244 intermediate stock which has a gamma of 1.0, so his resulting density increased by .30 ND times the gamma of 1.0 which yields .30, or 12 printer points, since .30 divided by .025 is 12.
- some wisdom from the old optical printer days in Hollywood.
I think I might be able to use textures like these to simulate wet plate collodion effects.
Sometimes thing developed in 2D translate well to 3D.
I had been playing with Alien Skin's Eye Candy plugin - in particular, the Swirl filter, which produces lines that flow either along or across luminosity gradients (more here) when I realized it might be useful to apply the filter to 3D surface curvature-based gradients produced by ZBrush in order to make textures that better respond to the underlying topology of an object.
I made a test object in ZBrush that had holes and branching, and exported various automatic cavity and smoothness masks from ZBrush to which I applied the Alien Skin Eye Candy Swirl plugin (combining both across and along gradient patterns) and wound up with something a bit like tree bark that seemed to understand the shape of the object.
Here's a nice write up of some of the work on Alien Skin's blog.
I did something remeniscent of this using custom 3D software for the film Little Monsters (1989) -- at that time I modeled a particle system after the rendering technique used in ILM's Star Trek II 'Genesis Effect,' which was discussed in a SIGGRAPH paper at the time.