CS527 Computer Animation Final

"Seeing Red" Computer Animation Critique

Episode 7 of Season 2 - Syrtis Major


This assignment is exaiming a sequence of computer animation program and select a short period of it to critique. The animation chosen here is "Seeing Red" animation program episode 7 of season 2, "Syrtis Major". The sequence that I choose is from 4:54 to 5:44. Beside this sequence of computer animation I also wrote down some of my overall ideas about the whole computer animation.

The Time Line of the Selected Animation

4:54
 Landing. Engine flames
4:57
 Parking act
5:13
 Vehicle moving out
5:14
 Should have trembling trails
5:21-5:31
 The whole vehicle moving
5:36
 Smoke and dirt by the wheels
5:44
 Lighting of the vehicle

Landing and Spaceship Parking

landing parking
The engine flames looks like using particle system as flames to produce the engine thurst powering effect. It looks working but my only problemis that flames in the front seems not as in detail as the ones in the back. In the later spaceship parking act shot, I think the animator should add a little subtle movement (like the action the pilot did with the flight control, a little bit moving up and down) to emphasis the weight of the craft. It should be fairly simple as implementing a damping translation up and down but it could add the sense of reality greatly.

opengate moving out
In the next shot, the cargo bay's door opened to let land rover vehicle moving out. The bay gate movements are so weight-less.  And the hinge rod to hold the gate are really too simple and small as if the gate is really light, but in fact it should not. Then the trails for land rover vehicle driving down to the planet's surface should move/shake a little to show the heaviness of land vehicle too. As the land rover vehicle moves out, a trembling trail could help showing heaviness too. Once again, this could be some fairly simple and easy implementation but these subtle movment could enhance the sense of reality. Looking closely, one can find that the two front gear of the land rover vehicle do not even contact the trails in the shot of vehicle moving out of the cargo bay.

Moving Vehicle

WHEEL veh inside intro better
When the vehicle is moving, in the shot inside cockpit the camera is shaking slightly up and down to show that this is a bumpy ride to the mine base. Also characters' subtle vibrating movement (leaning forward and back) also looks great letting you feel that this is a really a bumpy ride to the mine base. But when the camera shot moved outside, although the dirt and smoke after the wheel attempt to make it looks like really moving, but the vehicle seems to be in stable and smooth movement with the wheels not translated even a little bit with the bumpy surface of the route. I compared the land rover vehicle's movement with some shots in the introduction of this episode. I found that the animation of moving land rover vehicle lacks great amount of details compared to that in the introduction sequences. Maybe that is because audience will watch the introduction sequences over and over again but this episode's land rover vehicle animation, so the animators just put more details into the sequence animation an sacrifice the one in this episode? I assumed they might be using some techniques like bump-mapping or "Parallax Occlusion Mapping" to reduce number of polygons of the surface of planet. So even you see there are a lot of rocks and debris on the surface, but when some object moving across it, even they have a physical engine in the final rendering, the moving object is just moving on a big local smooth raytraced quad. Those sub-objects of the land rover vehicle will not have any interaction with the bumpy surface of the planet. So I think one possible way to make this sequence work could be, let wheels move a fraction amount of magnitude along the surface texture (as normal or height). This will make the movement of vehicle moving on the bumpy surface of the planet not only more realistic but more importantly consistent with camera views inside the cockpit.

Around the Vehicle

fake smoke smoke shadow
The smoke and dirt by the wheels seems stable, and may not really using a particle system in all shots, because in some shots the dirt cloud just stays with the vehicle's wheel all the time. But in the later shot, they put several particle system originated from the wheels. If you look closely in the second screenshot in the above, you will find that the dirt smoke was emitted as if there are several nozzles attachted the wheels, which is really un-realisic. And this particle system seems to be initiated from the second of the starting of this shot. It might be a better idea if letting the particle system start first for a while and then starting the shot. This will make the transition less awkward, since the vehicle was already moving across the ground and dirt before this specific shot. Also using a global scope and adding particles mass should have the dirt smoke inertia and become a little behind the vehicle and moving in rendered scene. Because it seems that the dirt smoke is "a part of" the land rover vehicle, but not a part of the planet bumpy surface ecosystem. They should generate another types of dirt smoke that is the turblance of the environment and it will be part of the environment and stay at some location and settle down instead of moving with the land rover vehicle all the time. Showing some distant dirt smoke and turblance in distance could make the vehicle riding scene more appealing as if the vehicle really interacts with the environment, not only spacially but also temporly.

The lighting of the vehicle seems not effect the surrounding environment at all. I think they might just model it as few thin light color transparent cones. Also in the original animation, the land rover vehicle's movements is just like floating with wheels doing their own rotation, the wheels should be shaking when moving across the surface of the red planet.

The shadow at the right side seems come from no where.

Floor and Wall

wall and ground
My first and biggest impression about this animation is that, all surface and material are so shiny and reflective! Not just the floor but also all the walls, character's skin, armor and machines. Their surface are really shiny and smooth! Which I think is not really fit into the scene and environment. The animator should try to reduce the reflective parameter of the material parameters in all surfaces. And in the sense of time spent on animation rendering, using less reflective material in the scene should reduce the complexity of raytracing and also the time needed to get the result animation. It should be a good thing for a program that needs to be aired in tight time schedule. If a realistic and complex surface is needed for a timely fashion, one can try to using "Parallax Occlusion Mapping", which will introduce more characteristics on the surface of the surfaces without bringing a lot of polygon complexity that will increase the time to rendering the scenes and whole animations. With proper graphics processing unit hardware, in some scene "Parallax Occlusion Mapping" can even achive real-time frame rate.

"Parallax Occlusion Mapping" is a technique of using big flat polygon to replace detailed geometry model having great number of polygons. In scenes that great background details are needed but a single texture rendered wallpaper not sufficient, maybe due to required view-dependent result, "Parallax Occlusion Mapping" can be used to achive nice detailed effects with little rendering time efforts. For example, in [ref. ], a 1100 polugons with parallax occlusion mapping will take less than 14MB in size and achive 255 frames per-second on ATI Radeon hardware. But if the whole model is represented in original geometry, it will take 1,500,000 polygons and 45MB in size, only makes to 32 frames per-second on the same hardware.

Facial Expression

weird
All characters seems not having too much expressions. Even they do, their expressions are rendered from direct facial motion capture dataset. It seems pure motion capture or skinning was used, there is no human muscle model involved. And while people talking, their eyes blink much too less and un-real, and their eyes rarely move at all. To remedy this, building a generic human muscle and kinematic model could provide some good results.

Human Skeleton, Limbs and Muscles

skefacial
The motion capture of body movements seems working seems OK. But in some shots, it seems that the motion capture sensor is not properly placed/attached to the control points in acting characters. That makes some postures and moves looks awkward, for example the first screenshot above. And the character animation seems not considering too much complex human limbs and muscles modeling. That makes the characters look like puppets with cylindrical tubes as limbs, and looks funny. And their approach almost totally not separate character body parts moving from clothes deformations. I think they treat the clothes and body as the same thing, which means there is no such concept as clothes and body parts. For example, if you look at the second screenshot above and related sequence, you will find that the man's suit does not wrinkle or fold at all. Clothes are just as the skin to the character. Also, I was wondering whether they are hav two separated animator teams, one group working on the head/facial motion capture and animation and the other group working on the body gesture motion capture and animation. The reason that I made such assumption is that from the over all look of a character, the heads size seems not propote to the body or limb size at all. Take "Captain Scarlet" for example, from his head and the character's feature, he should have more stronger arms and body features, but he seems to be just every other characters, having slim limbs and body figure, which really does not look right.

Character's Weight

jump
In some shots, like 15:33, the character movement (jump and grasp the rod on the top) is supposed to be a big motion action. But the way they present it makes audience feel the character is weightless. With some minor details in accessories the shot could be made to be more realistic. For example, they can increase the amplitude of the character jumping to grab the upper rod, so it will not be like the character 'hover' up all in the sudden. Making the action starting from a short run will help too. Also it will also help if the upper rod bend a little bit or with additional metal sound showing something with weight apply some force on it, instead of two object just translate and glue together and there is not really interaction between them.

References

2006 Fall by Julian Yu-Chung Chen