TAPESTRY: The Art of Representation and Abstraction
Animation Delineation
The Problem:
Motion is central to animation, but the screen is almost invariably a static (unmoving) display. Further, we don't really want to identify each and every camera position and object position over time. How is the motion to be defined, visualized, and edited?Further, motion breaks down into two distinct types: camera (or viewer) animation (dolly, pan, zoom) and subject (or moddel) animation (bouncing ball, opening door, operating automobile engine, walking person, etc.) Both of these worlds offer rich opportunity for computer graphics, and present some particular challenges.
The Solution(s):
Story Boards and Key Frames
Generally we want to define "end-points" for animation, in much the same way we define a line segment by it's end-points. Only, animation operates in a four-dimensional space--the three spatial dimensions plus the time dimension.
This general problem relates to the others, so lets look at it first. The motion picture industry has already addressed this problem in deciding how to shoot & sequence movie scenes. They use a "story board" to describe the sequences. The story board is essentially a comic book, a sequence of images which convey camera angle, action, etc. Each one is an icon for a brief sequence in the film. Together they describe the movie.
Most computer animation schemes involve manipulating "key frames" in some way. Key frames correspond to the story board images. The system must know how to get from one key frame to another. The general name for this process is "'tweening" (short for "in betweening").
More on that in a bit. First, consider what might be changing in those key frames.
Camera motion
Modeling programs have mechanisms for specifying viewing conditions. If we could define two distinct viewpoints for the program, and leave it to interpolate the intermediate viewpoints, we'd have the beginnings of a key-frame animation environment. By changing the number of "tweened" frames from 1 (stepping directly from view #1 to view #2) to a larger number (depending on how long we wish the sequence to last) we can achieve the desired time compression, giving both the control and the abstraction necessary for effective editing. However, "tweening" isn't as easy as it might seem, see "The Problem With Motion," below.
Subject motion
The alternative to moving the camera around the data is to move the data in front of the camera. Every computer modeling program has mechanisms for editing the geometry of the model. If these edits (generally not deletion or addition of entire objects, but manipulation of existing objects, like MOVE and TURN) can be saved as an "edit script" and later applied over a number of frames, the effect of motion will be achieved. If we consider the camera to be fixed, then we can edit the position of data objects in each "key frame", rather than the position of the camera, and then 'tween this manipulation when creating the animation. By varying the amount the ball moves up or down in each key frame, we could create the illusion of a ball bouncing in front of the camera.
In fact, we could let a key frame keep track of both camera position edits and data position edits. This way the key frame data describes both subject and camera motions. (Note that it also ties them together, so that making the camera circle the bouncing ball at a faster rate, while the ball bounces at a fixed rate, becomes more difficult because the amount of motion between frames needs to be different.)
Other "motions"
The movie at right illustrates two other kinds of "motion" which might occur in an animation. On the left side, in the back, you see an object with an "animated" or "rotoscoped" texture. On the spinning (moving) object in the right foreground you see another example of one texture blending, or "morphing" into another. You can similarly imagine, perhaps, moving light sources, variable transparency, and so on lending material to your "story".
Synthetic Actors
Researchers (including some at the University of Washington, check out
this movie) are exploring ways in which to further simplify defining motion by creating abstract motion models for gaits ("walking" vs. "running" vs. "limping", etc.). They cost a lot to develop, but they don't get residuals, don't throw tantrums on the set, and don't wrap their cars around trees. They aren't very good actors yet, but they're learning.
Last updated: April, 2014
