| This is a demonstration of a proprietary facial
gesturing and animation system designed and implemented by me for the 3DS Max Environment.
It is presented as a proof of concept and should not be considered a complete solution at
this time. The primary focus of this system was to create a powerful and versatile
gesturing system that would allow data to be easily shared between characters.
Traditionally, morphing or bones are used to do complex facial animation. Although
morphing is a powerful tool, I felt that the time and overhead needed to create all
the morph targets for multiple characters and the fact that it is possible to break the
volume of a character when mixing multiple morph channels together wasn't the best
solution for facial animation. Using rotating bones allowed me to share data between
characters easily and since bones are independent of surface topology, they didn't require
the time consuming task of creating morph targets. A bone system could be setup very
quickly however it didn't offer the fine control of shape and volume that morph targets
provide. So, I set out in search of a better solution. The hardest part of this was
developing the idea. The actual implementation was fairly simple. Let's have a look. |
| The system is
based on groups of nodes that slide across hidden ghost surfaces simulating skin sliding
over muscle and bone. The red nodes shown in figure 1 are the drive nodes
that control the system. These nodes are constrained to slide across the blue
<ghost> surfaces. These are the nodes that the animator interacts with. I setup 22
drive controls to control the muscle node groups of the face. I wanted to keep the system
fast, fun and easy to use so the idea here was simplicity. The next step in this area of
the system would be to add a method of capturing a gesture such as a vowel sound or frown
for later use. Now lets look at the core of the system. |
|
|
|
In figure 2 you see what's under the hood. The yellow nodes shown here are
driven by the red nodes in figure 1. These arrays of nodes are linked to each other in
subtle and sometimes complex ways. Each node is aware of the state and position of its
neighbors. I also provided direct access of these nodes to the animator for fine
animatible control of the system and to allow the system to be tuned for use on a
different character. Generally, the drive nodes in figure 1 were all I needed for great
results. The next step here would be to automate a lot of the setup process by adding a
front end interface for group node reactions. |
|
|
|
| Figure 3
shows what the animator sees as they work. On a single PII 400 I was able to animate
real-time. Theoretically the system could be hooked up to a MIDI board system to
facilitate real time capture of animation. Since mesh topology and resolution remain
independent of the facial system they can be edited at any time. |
|
|
| The final result (figure
4). Soft fluid motion and consistent volume. To view a demonstration of how the
system works as MPEG (3.7mb) click on the link below. |
|
|
|
|
|
