Buxton, W. (1999).Where
IT's at. CGI Magazine, December, 1999, 16-17.
Where IT's at
Bill
Buxton
Chief Scientist
Alias|Wavefront
buxton@aw.sgi.com
www.aliaswavefront.com
September 27, 1999
Introduction:
My sense is that the computer
graphics industry is at a turning point comparable to what happened in
the early 1980s, when it started to emerge from the domain of flying logos,
to images of nature and character animation. I see us moving into a period
that will have the same excitement we experienced the first time we saw
Lorne Carpenter’s film of a teapot in the fractal mountains, "Vox Libre,"
or Turner Whitted’s beautifully ray-traced images of crystal balls on a
checkerboard table cloth, or chess pieces on a fog shrouded board.
To anticipate a turning point
is one thing. To characterise it is another. Yet that is what I am going
to try to do, albeit briefly.
Coming from the music technology
field, I have always found that what happens in computer graphics is parallel
to what happens in computer music, but delayed by a few years. This delay
does not imply that computer musicians are smarter than those doing computer
graphics: rather, it is a reflection that computer graphics is significantly
more demanding computationally. Once recognised, this parallel can help
us anticipate and understand what is to come.
Music Lesson # 1: Desktop
Performance Animation
As computers become more powerful,
so does our ability to deal with complex models interactively. For example,
we can now scale, translate, rotate, mutate and animate things on desktop
machines in ways we could only dream about one or two years ago. And we
can do so with them fully shaded. Now, once we have reached this stage,
we are confronted by the following questions:
When you manipulate
a complex model in real time, what is the difference between manipulation
and animation?
The answer, in my opinion, is
simple:
It just depends
on whether or not you have the record pedal depressed.
Like we have done with music
software, we are now at a point where we can "perform" animation at the
desktop, i.e., do "desktop motion caption" or "desktop performance animation."
This will lead to concepts like "multi track / pass animation," where we
build up and refine the animation layer by animation layer. In addition,
we will refine it by something that might be called "motion mixing." In
doing so, I suggest that MIDI controllers and other input devices previously
seen only on musicians’ desktops, will increasingly be put to use by animators.
One result of this change
will be that animation partially moves out of the back room and into the
front suite. This is a result of increased interactivity, which means the
3D animator will be able to work side-by-side with the client more to refine
his or her work.
Music Lesson #2: From Synthesis
to Sampling
In music terms, the computer
graphics industry is in the mid-1980s. Why? Because that is when a company
called Emu brought out one of the first affordable samplers, The Emulator,
and forever changed the music world. So how does that relate to those of
us in computer graphics?
Think about this: until now,
all of the computer graphics companies (such as Alias|Wavefront, Softimage,
Kinetix, etc.), in music terms, have been synthesiser companies.
That is, they created tools to synthesise images. However, one of the biggest
changes over the past 12-18 months has been the introduction of a broad
range of technologies that enable us to capture data from the physical
world, i.e., "sample" it, and use it in our work.
Let’s look at this in more
detail. Now the reason I use the term sampling rather than scanning
is because (a) it helps establish the music analogy, and (b) "scanning"
carries the connotation of sampling just objects, and there are many more
things that can be sampled. By seeing these in the same light, under the
same term, the emerging pattern becomes clearer, I hope. So, here are some
examples of things that will be sampled more and more:
-
Textures: 2D scanners
and digital cameras enable us to sample textures that can then be mapped
onto 3D data, regardless of source.
-
Camera Location: Camera
extraction software such as MayaLive™ enables us to "sample"
or extract camera position from live film footage.
-
Animation Curves: Motion
capture systems can be thought of as a means of providing sample animation
curves, rather than of a way to synthesise them.
-
Object Geometry: There
is an emerging range of new technologies for scanning 3D objects. These
range from contact scanners, to laser scanners, to photogrammetry. Prices
are dropping dramatically, and this is one of the areas where, over the
past two years, there have been some of the most innovative and newest
companies emerging.
-
Lighting: We are starting
to see, from Synapix, for example, suggestions that photogrammetry
(i.e., measurement from photographs) is going to enable us to extract lighting
information from a scene through a computer analysis of shadows and highlights.
This technology, when deployed, will mean that the photographic image itself
will become a significant part of the renderer’s "user interface" when
trying to match CG lights to a live action scene.
-
Sets/Locations: The use
of LIDAR technology, such as that practised by the team at Panavision;
photogrammetry, such as that practised by SynaPix; and the emerging
depth-sensing video cameras, such as the "Z-Cam" from 3DV Systems, are
enabling computer graphics artists to work from a 3D model of the scene,
without having to reconstruct the scene from scratch. The implications
for those doing set enhancement, or inserting 3D characters into real world
scenes, are immense. We are about to move from the absurd situation where
we have 80-200 layers of compositing (in pursuit of getting 3D effects
using 2D technology), to a situation where we do 3D work in 3D, thereby
allowing us to focus on content, rather than administering hundreds of
assets.
Going forward, digital imagery
will, like today’s music, be composed of a mix of sampled and synthesised
materials. Hence, to stay current, the CG packages of the future must support
both types of material in a seamless and integrated manner. Recognising
this and other trends early is one of the key reasons that Alias|Wavefront
invests so much in research. Once recognised, this drives many new partnerships
that we are developing, as well as R&D initiatives, such as applying
our automotive design point-cloud processing tools to our entertainment
business.
Music Lesson #3: Towards
Physically-Based Modelling
Finally, we are starting to
see signs that CG is moving into the current "hot area" of sound synthesis,
physically-based
modelling (PBM). PBM is the use of equations approximating the physics
of an object in order to synthesise it, as opposed to creating (synthesising)
or copying (sampling) its appearance. Dynamics and inverse kinematics are
early uses of PBM in computer graphics. Marking a change on this front,
we are now seeing a new generation of graphics engine emerge, represented
by the new game engines such as the Sony Playstation 2. This will evaluate
dynamics and IK in real time, rather than precompute them.
PBM will go even further
than this. A good example is Jos Stam’s research (of Alias|Wavefront),
which appears in the Proceedings of SIGGRAPH ‘99. This work develops
a set of formulae that enable one to approximate 3D fluid dynamics (such
as the behaviour of smoke, fog, or liquids) in real time on a desktop machine.
The effects are dramatic, and are just a hint of what is to come, from
both us and others.
Conclusions
We live in interesting times.
For the past 5-10 years, what was a 2D and 3D graphics package was pretty
well agreed upon and understood by all. Progress from year to year was
mostly in the form of new features, and improved price-performance.
But I think that is over,
and that the field is about to be redefined. The changes happening today,
and not just those mentioned above, are going to shake up the industry.
We will have to look at how we have been doing things, and at our assumptions.
We will have to ask, "Have we been doing things this way because it is
the right way, or because it was the only way that we knew how at the time
we started doing it?"
My belief is that increasingly,
we will realise that the structure of the industry is largely an artifact
of the early ‘80s, when trends were set. We did things in a certain way
because that was the best we could do at the time. Well, times have changed.
For the first time we can begin to ask, "What is the right way to do things?"
and have some hope in the ability to carry it out.
Asking such questions is
easier said than done. But the future belongs to those who do. Make no
mistake: we plan to be there stronger than ever.