#Jurassic Park
22 messages in this thread
Just saw Jurassic Park today. If anybody's ever doubted the
computer's ability to realistically portray the world you better be
prepared to be amazed! I won't ruin anything by telling stuff, but
you need to see this movie! Just remember…most of the scenes with
dinosaurs were CGI.
I read in Computer Graphics World that they suspect that Metaballs
were used to do some of the animations. Have you heard any such
talk? Aparently Metaballs are incredibly useful for creating organic
shapes but very difficult to apply maps to. If you have not read the
article it might be interesting to you. But then, it was a fairly
cursory treatise. Any luck we'll get something like this on our
platform?
Jim,
I saw the metaballs article as well. That technology sounds pretty cool. I
don't think the Amiga platform will see any metaball modelers any time soon.
It would seem that spline and polygon modelers are too different in nature to
accomodate a metaball addin. In other words, a modeling package that supports
metaballs would have to do so from the ground up, IMHO. And that will take
some time!
That would definately be cool, though.
>I saw the metaballs article as well. That technology sounds pretty
cool.
I have only seen a animation done by some Japanese which was shown in a
Siggraph tape and they quickly described it but it wasn't that clear.
Could one of you guys explain what is so special about MetaBalls?
<<< Tony D. Tracer >>>
I'd suggest that you get Kawaguchi's book, "Growth Morphogenesis". He explains
what Metaballs are and how he uses them. The system that rendered the Metaballs
and the software appear to be proprietary. I wish Imagine had some way of or-
ganically morphing smooth shapes – letting them grow in a fractal pattern over-
time.
Metaballs are defines by a point that carries with it radial information and
has a threshold at which it will fuse with another metaball. They could be
described as behaving like drops of mercury for some types or like clay or
putty for other types. Super Metaballs are defined additionally by allowing
the modeler to narrow the direction if which the fusing takes place. This is
useful for modeling, say, a hand. You want the finger joints to fuse but you
do not want the fingers to meld into a mitten-like object.
They can create very complex shapes that would be practically impossible with
polygonal or even spline-based modelling. They pack into a small file usually
about one third the size of a poly file.
To give you an idea as to their versatility, it takes about 43 metaballs to
make a beautifully designed human figure. Unfortuneately, the problems with
rendering them is that they are mostly impossible to map. How can you map
objects that fuse?
If you want more info, I got this and more from the latest Computer Graphics
World.
Jim Roof ?exit (let's try that again)
Simply,
Metaballs are kind of like adding lumps of clay to an object.. if you need
more mass in an area you glob a metaball in and it blends in seamlessly
into the rest of the model.
JG
I bet the renderer is pretty specialized too, although raytracing with
only sphere primitives is very fast, it would not be optimal unless
it knew what to expect.
Hi Mark,
Well, I think it'd be *possible* to use metaballs in a poly modeler…!
Metaballs and blobby modeling both fall under "implicit
modeling"…using algebraic expressions (instead of a listing of poly
vertices, or a parametric patchmesh description, or quadrics) to
describe a shape. The "sphere of influence" that Jim Roof describes
is a special case of this, but not the only way it can be used! The
hard part is figuring out the equations needed for a certain shape,
and I think that may be why most of the action is on the blobby
methods now — it's easier for designers to interface with.
But it's tough to render these equations. One popular way is to first
polygonize the data, and then send it out to a normal renderer.
Paradoxically, it's easier to fit poly vertices to a specified shape
than it is to fit a spline patch to it — the tradeoff is lower
fidelity vs. quicker development. But blobs->polys is being done
today! (And isn't this "zone of influence" idea similar to the way
that bones deform vertice positions?)
It seems metaball computation is a bear, though…takes significant
time even on a workstation. There's already a PC program (it's either
Poly-Ray or POV-Ray) that uses Blinn-style blobby modeling…haven't
fired it up yet myself…more info in GraphDev. And there's what
sounds to be a hot class on Implicit Modeling at SIGGRAPH this year,
with many of the folks who've done the basic research.
Odds are that we'll be able to use this stuff "soon" — cool! Anyone
have more info?
Jim,
I saw the program being used at SIGGRAPH in Chicago last year. They had a
small booth with hardly any people around. I managed to get a "tour" of
the program. reeal nice stuff, and interesting concept. After a while you
start getting the feel of working with clay balls…
=Louis
On AUTOPILOT over Chicago Stadium, GO BULLS!
I believe Electrogig's 3D GO Software for the SGI supports the use of
Metaballs.
I've heard a few different things about it. I am going to a gathering
this Saturday to find out from ILM how they did a lot of the shots.
Should be pretty interesting.
I wouldn't be surprised if Allen & Stuart came up with a better way than
Metaballs to do organics in the future. Bones are better than freeform
deformation….
JG
John,
I am not a mathemetician, but I have been modelling organic shapes in LW and
from the various experiences that I have had, it does seem that there could
even be algorithms applied to the joining of polygonal masses. It would
require the movement of a good deal of points to create the common vertices and
still allow for good smoothing to take place. This is always the rub.
It seems that the shapes could be intersected, then based upon a
user-defined radial distance from the center of the intersection, the
points that fall within this distance would be moved as if being
stretched along a rubber band until the points slowly are displaced
over the distance until they become a shared point. I know this is
problematic, but it is basically what I am doing in painstaking
detail at this moment! The original primitives actual would change
shape, but for organics that would be OK as long as the basic linear
structure was maintained. I'll shut up now!
John,
ILM. Ooh. Ahh. (castigated bow).
Have fun. Hey, ask if they got any Amigas there, OK?
… Nate, Emerald Video.
I Animate – because – I am a Nate!
What's the movie supposed to be about?
=Louis
On AUTOPILOT over Chicago Stadium, GO BULLS!
Jurassic is based on Michael (Andromeda Strain, et al) Crichton's book of
the same name. Basically, a way has been found to clone dinasaurs from
their DNA strands found in mosquitos frozen in amber. A very rich
man/company builds a 'theme' park on a remote island full of these cloned
dinos and a few experts are invited to approve the park and…..things
happen!
If the story doesn't grab ya, the movie is a must see for any body
interested in computer graphics. The majority of the shots were CGI.
JG
The book is outstanding… start raytracing a complex scene of about 200
frames and by sunrise you'll be done the book.
WAnd after the last page you'll realize that you had a flipped polygon. But
that's OK. You're probably on about frame 15 by morning anyway.
That's about how long it took me to get through it! Although, by the
time you're done with the book, you realize it was written kind of as
a screenplay… I loved it tho.
Several of my other astute friends realized that the book was pretty
much written as a movie. The pacing, the high degree of visual
stuff… Have you read Sphere or Congo yet? Also excellent Crichton
books… Sphere is ALOT like the Abyss, at least in scenario and
characters.
Not yet, perhaps when I get some free time… RIGHT!!!!
Thanx for the min review of Jurassic Park.