figure 9. screenshot of the Plushie system and plush toy designed with the system.
them to computer representations. In
the case of 3D modeling, we learned
from sketching activity on real paper.
For 2D animation, we learned from
children playing with a real plush toy
using both hands.
Instant feedback. Instant feedback
is critical to real-time interaction. It
allows for graceful learning through
casual trial and error while supporting creative exploration through rapid
experimentation. To provide a rich and
comfortable user experience, three opportunities for executing computation
should be used: One is computation during mouse dragging, a computation that
must be very fast (on the order of 0.1 seconds). The second is computation right
after a mouse click or dragging; it can be
somewhat slower (about a full second).
And the third is running a computation
in the background while the user is looking at a result. The sketching interface
is effective because it gives a system the
opportunity to execute a heavyweight
computation right after the completion
of a sketch (mouse release) that would
otherwise be too time-consuming during
a mouse drag. In 2D animation, the system computes time-consuming matrix
factorization when a pin is added or removed, applying fast back-substitution
during dragging. The clothing-manipulation system exploits idle time to refine
the cloth configuration.
Right target task. System designers
must choose and focus on the right tar-
get task to achieve the first two goals.
Developers try to address a range of
tasks, overloading the interface with
too many functions, as in professional
systems like Maya. In theory, including
more functions could expand the range
of user options but also require inten-
sive training and reduce what casual
users are able to do in the system. Care-
fully limiting functional scope, design-
ers provide an optimized interface and
algorithm for the task in exchange for
losing some rarely used functions. Ted-
dy is designed for rotund models (such
as stuffed animals), freeing users from
having to specify depth information
each time. The clothing-manipulation
system simplifies the interface and ac-
celerates the computation by focusing
on the cloth on the surface of the body.
System designers are better off tapping
user creativity than constraining it with
many predefined functions. A simple,
well-designed interface allows users to
apply their imaginations to complete
tasks beyond the system designer’s
original assumptions, as in terrain
sketching with Teddy.
In addition to improving the tools discussed here, we plan to work on other
aspects of computer-graphics authoring in the future, including two notable problems:
Designing interactive behaviors.
In-teractivity is an important aspect of
computer-generated media. Not only do
users passively watch predefined imagery, they also interact with computer imagery (such as by poking a character) to
observe its response. The systems we’ve
introduced here are all interactive as authoring tools, but the content they produce is noninteractive; 3D models and
2D animation created this way do not
respond to user input. End-user design
of interactive behavior is an exciting but
challenging research direction.