STAR BORES

• Modeling:

  • Evangelion: Modeling was done in both 3D Studio and Data Explorer using primaritly CSG operations and simple transforms found in both packages. The model of Evangelion is composed of several seperate parts that are combined to form a single model. Everything in the model is hierachical, from the arms and legs, to the two little eyeballs. The "basic" units, such as the upper arm, lower leg, shoulder blades, etc, were modeled, scaled to their correct relative sizes, then scaled again by [-1,1,1] to get a mirror image of the object to be used on the other side (all elements were modeled as being on the right-hand side, relative to the model).

    The coloring comes mainly from colormap modules. The objects are marked as positions, then the minimum and maximum values (from a statistics module) are passed into the colormap. The marked data then has the color applied to it from the colormap. Some objects are a solid color simply because we wanted them to be.

  • Death Star and Trench: These elements are made from two hieghtfield created with a paint program. The images were read into DX, and the color found at each pixel is used as a hieght for that location. There are two seperate maps: one for the surface of the Death Star and floor of the trench (the map used on the floor has been flipped over), and one for the sides of the trench. For the parts of the trench that are very far away (realatively speaking), fake a fake floor and walls are used to speed up the render time. The fake elements are both made of a 2x2 planar construct. The heightfields (real and fake) both are scaled by very large amounts to make the trench appear to converge in the distance. The fake floor and walls are somewhere in the neighborhood of 70,000 units in length.

    In case you have defective eyes (like me), the sign at the end of the trench reads as follows:

    Danger!
    Death Star Weakness!
    Keep hands, feet, and Proton Torpedos out of Exhaust Port!

    Nice of the Imperials to place a safety sign, eh?

  • Tie Fighter and Hand of Death: The hand is done solely in DX. The fingers scaled tori, while the fingertips are spheroids. The bulk of the hand is a scaled sphere. The Dark Lord of Sith's craft is made partly (i.e. the wings) form cheap, prefabricated parts found in Avalon (ahh...how I miss the land of the faerie and the days of King Arthur). The body is made of DX objects. The wings and body are seperate so that the former can be forcefully detached from the latter, while allowing each part to be animated seperately.

    The explosion is just cool...okay, I'll tell you how we did it. It is essentially a marco that takes an object that has been scaled to one unit in size, the output is then transfored as the original object is/would have been transformed. In addition, the macro takes a start value, a stop value, and the sequencer. When the sequencer reaches the start value, the explosion is created and begins to expand. It is gradually faded out, along with the original object, as the sequencer counts towards the end value passed to it.

    The animation for the hand is composed of a simple translation and a bit of hieracical modling in the fingers.

  • The Dragon: If only the Architects could do something this cool... The dragon was done for a previous project (lab 3 if memory serves). It was altered slightly to fit our evil aims, but is mostly unchanged. The wing motion is sinusodal, and the claws graps the shoulders of Unit-01 correctly.

  • Pacman: Genuflect and praise the Grandfather of all video game heros. Pacman is two hemispheres that rotate according to time. He simply munches along and eats small (or not so small), round things. This time however, it didn't agree with his stomach very well.

• Evangelion Animation: (save the best for last ) The animation of Evangelion is based on several sources: a picture of Evangelion running full-tilt, pictures of Brazillian soccer team, literlly running around in the Cornell CS department's Undergraduate Lab (a behaviour that prompted many strange looks from the other people in the lab at the time), the Ranma 1/2 first season opening (there's lots of people running), myself, and many long hours trying to find a periodic funtion without doing really knowing how (I did it anyway).

  As the model is hiearchical, I based the motion around the angles between joints at nine different points: two for the elbows, knees, and shoulders each, and three at the hips (two for the legs, one for the rest of the body). The rotations at the shoulders and elbows are in phase with each other so that when the arm is in from of the body, the arm is more bent than when it is behind the body. In addition, the lower arms twist inward when appropriate to give a more fluid and realistic motion to the run. Both the upper arm and upper leg rotations follow a sin(x)^0.75 pattern asthis allows for more "loiter" time for the limbs at the extreme ranges of motion. This had the effect of slightly exaggerating the running and also give more of a dynamic look than a simple sine or cosine function would do. The lower legs are slightly out of phase with the upper legs for two reasons: it looks better, and is much more realistic in terms of how running actually works. The slight phase shift allows the leg to be tucked while it is being brought forward, and extended while pushing away from the ground.

• Post Production: All the scenes were generated by DX (about 1 DX file per physical scene) and individual frames were output as tiff files. the .tiff files were uploaded to a Pentium Pro 233Mhz with 64 megs of mem and aprox 4 gigs HD space. There, the frames were assembled into an .AVI movie using Adobe Premiere. Then, ULead Mpeg converter was used to encode the AVI into .MPG format. (note, no premiere effects were used in this movie. Adobe Premiere was used only for assembling the frames and making the movie since doing all the seperate frames by hand is a real pain in the [censored by the government of Surinam].)

"Where do we go from here?"

There are a number of things, specifically having to do with the animation, that can be improved upon. Another equation that moves Unit-01 up-and-down in phase with the running would add to the realism, as would adding more pivot points to the model. Such points, perhaps in the lower torso, neck, and several in the spine, would allow for a more fluid motion. A rather unhappy side effect of doing this would be a rather complicated set of equations, and with only nine pivot point now, it gets rather messy as it is... The resolution of the hieghtfields could be improved, perhaps using a special filter. The original maps are intentionally boxy to give the correct surface effect (watch the movies and you'll see what I mean).

Should you have questions you can e-mail Jesse or Joe

Page by Hawson -- Updated 4/21/97 at sometime waaaaay past your bedtime.
This page created with PICO.