Saturday, December 18, 2004

Blogalog 1.0.2: Skin

We'll get to the latest questions on quantum computer architecture shortly. Meantime, a CG question:

What's the latest on human skin? I've heard (from you) that it's extremely tough because it's partially translucent; I assume the way it stretches, folds, and wrinkles are tough to get right, too.

What's the scoop? Is anyone making significant progress?


At 6:06 PM, Blogger mr.wook said...

The major topics in skin are in near-common use
today. Subsurface scattering (a major factor
in the appearance of translucent substances, be
it milk or skin) has gone from Siggraph papers
to production implementation in the last couple
of years.

Any curvable/folding surface (skin, fabric, etc.)
remains tricky, the `correct' time and way to crease
still generally requires a lot of manual
intervention when the surface does the wrong thing.

There is another problem: What looks best?
Whether you are shooting for drama, or sheer cool,
the result of a physical simulation rarely produces
results that are cinematically optimal. Some
of this is expectation when reality doesn't
match. (I am actually watching Matrix 3 while
typing this).

And there is always the backup solution:
Call upon the three great allies of computer
graphics: speed, distance, and darkness.
Speed (esp. with matching motion blur) minimizes
the eyes ability to absorb detail;
Distance -- the smaller the CG object is in frame,
the smaller it's flaws;
Darkness -- Why do translucency whe there isn't enough light to scatter?
For _The Hulk_ ILM generated an insane number of
layers for the Hulks skin to produce varying levels
of detail, including dirt, wounds, etc. Each
of these layers required shaders, textures, and
other significant support. If only we had
a quantum rendering GPU to draw on.


Post a Comment

<< Home