Critique of New Captain Scarlet
Computer Animation (CS 527) - Final Exam
Season 2 - Episode 11 - "Shape-Shifter"
5:12 to 5:33 and 5:53 to 6:00
Gerry Anderson’s New Captain Scarlet was a flourishing attempt to take a puppet show, remaking it using Computer Generated Imagery and calling it “Hypermarionation”. I chose to critique the following portions of the Season 2 episode “Shape-Shifters”: 5:12 to 5:33 and 5:53 to 6:00. (Henceforth, I shall refer to this time frame of the clip as 0 to 30 seconds, for the sake of convenience).
Click here to see the clip.
A Mysteron replica of Captain Scarlet brings the real (shot dead) Captain Scarlet into a forest and dumps him into a swamp. A hick saves the Captain’s body from being eaten up by an alligator.
Any animated video encompasses several techniques and ideas combined to form a sequence that looks as realistic as possible. Here, I try to dissect the methodologies used in the animation clip that I've chosen.
Models have been designed masterfully in the New Captain Scarlet. They have been kept to low polygons counts, probably to speed up rendering.
The car in my scene looks futuristic. Although I don’t know what purpose is served by having lights pointing upwards, it looks cool and shiny. Ferrari red colored two-seater cars with upward opening doors are pretty common these days, but that is a good point to start designing a car of the future. The other vehicles in the series are better and look awesome, especially the motorbike that can fly and the mars rover with lots of wheels and hyper-sensitive shock absorbers.
There is also a boat in this clip that belongs to the hick, which looks old and small. It gives you a feeling about the intentions of the people using it.
The people are designed pretty well. Most of the clothing is usually armor, which makes life a lot simpler for the animator / designer. The hick and his father have good physique. The hick looks dumb, something that might have gone in intentionally.
The faces look blotchy when you compare it to something like Dr. Aki Ross from Final Fantasy. Apart from the actual talent of the model designers and animators, rendering a face with very high detail can take up CPU and time, all of which translates into money. The faces try to show emotions to an extent, "and that's all I have to say about that".
Guns, for some reason, have not been an important thing for the animators. Although they never appear in this clip, the guns that the officers carry look like pipes with handles. Thankfully, the shotgun that the hick carries looks real.
The alligator has been modeled beautifully. A lot of attention has been paid to its rugged surface and scary teeth. Although it stays for a very short period of time, its presence and threat is felt.
Small things haven’t been paid too much attention. It is probably not worth it, because of the short duration of their appearance. These things are hardly noticed, like the motor of the boat exposes the polygons that went into its making.
Lighting and Shadows
The environment is mostly a flat texture and gives very minimalistic scope for animation within it; some textures (like the background in the one in with the boat) look very obviously like a landscape painting. It would have looked nice if there were a couple of actual tree models in the front with a textured background. Lighting and shadows are incorporated into the background texture, so there's nothing to bother about on that front. The animated objects are ray-traced and shadows are generated based on the position of lights.
This is where another problem crops in: In the first sequence (00:00 - 00:11), the background suggests that the sun (the light source) is at the top right hand corner of the scene. When the car moves in, the shadow is casted on the left hand side, which is perfect. Then, the fake Captain Scarlet gets out of his car and walks to the passenger side of the car. By the time he gets there, the shadow of the car has magically moved to the right. The two character models also cast shadows to their right.
Then, the fake Captain Scarlet carries the dead body to the river (00:12 - 00:16). Initially, the shadows are to the objects' left, and by the time they reach the stream, the shadow moves to the right.
This is a classic case of having more than one light source and not specifying their properties correctly. In an outdoor scene, one light should cast shadows (unless there is a lamp or something similar nearby), while the other lights should not. Setting a simple property on the light objects would have prevented such discrepancies.
Another way to avoid this problem is by using a single light source. Then, we may have to depend on the ambient light to light up the polygons not in front of main point / directional light. Using Radiosity is also good solution to light up scenes, but it would take an enormous amount of time to come up with so many frames.
Motion Capture is clearly the obvious method to use to implement walking and other human motions. This is a pretty straight forward modus operandi where you put sensors on an actor, record his motions and transfer the same on the CG Model.
In the clip that I’ve chosen, the fake Captain Scarlet walks more like a wrestler than like a Captain (00:04 to 00:10). He seems to be jumping up and down, and keeps his hands away from his body, both of which are not natural. This could be because of several possible reasons:
Acting: The actor has to get the feel of the character and has to walk with pride like a fake captain who has no guilt, as opposed to someone who’s scared of the sin he’s just committed.
Mo-Cap: The mo-cap sensors and the limited space available to carry out the actions could have prevented the actor from acting naturally.
Mapping: When mapping the model to the mo-cap data; the scale factor might not have been used properly.
Dead Man Falling
If you closely observe the two seconds of the clip where the fake Captain Scarlet opens to door to let the real (dead) Captain Scarlet out (00:10 to 00:11), the dead man magically lifts his hand and throws it out before letting the rest of the body fall on the ground. This is something that could have been easily avoided, simply by re-enacting the sequence correctly.
Mo-Cap while carrying objects
During motion capture, the weights carried by the characters in the scene have not been taken into consideration. The fake Captain may be strong, but that doesn't mean that he can walk around carrying a dead man with heavy armor on his shoulders without slouching, or slowing down. More on weightlessness will follow.
Key Frame Animation
Key Frame Animation is pretty straight-forward. The "key" frames are put in the correct, required frames and the engine will simply interpolate (move / rotate / skew / morph) the objects for the frames in between. But this can get tricky at times.
Face Shots and Lip Sync
There are a lot of scenes in this series with close up shots showing people talk. The characters talk a lot and most of the talking is shown up-close. The characters look like they are saying whatever is being said.
Using Key Frames is the best way-to-go to implement facial animation. Create a database of frames consisting of the various possibilities of lip positions, eye look and feel, and overall face expressions; put the right combination at the right place; let the computer interpolate the remaining frames; and viola - you have perfect lip sync and face expressions. There are algorithms out there that take in voice audio for the character and generate the lip movements automatically. I don't believe that any such thing has been used in this series. In my clip, there are only two words, and it has come out well!
The face model itself can be a lot more detailed. Some men (including Captain Scarlet) have a one-day-old beard, which is good. The teeth are not one long white patch; it can actually be seen as separate teeth, but in an attempt to over-emphasize this fact, they have made it look worse than the single white patch. The eyes, to some level, try to show the emotion of the person. For something that was done recently, it is not unfair to expect a higher level of detail. The skin tone can be much more realistic. The hair can be separate strands of hair (like in Shrek 2) rather than one big lump of black / gray / white colored hay. But then again, time and money are factors that would have restricted such eye-candy. Putting caps on their heads is a brilliant way to minimize the amount of hair visible. When you think about it, they could have had a character with a clean-shaved head - it would have looked different, and less work for the programmers / animators.
When the hick shoots at the alligator, the response to the three gun-shots has been animated masterfully (00:22 - 00:26). The alligator responds to the three shots distinctively. It would have been cool if a little blood sprouted out right away; it still looks realistic. The water splashing has complimented the alligator's motion each time the tail or the head beats the surface of the water. A lot of key frames must have gone into this small piece, and it is a job well done.
Just when it seems like it almost came out perfectly, the tire skips a few degrees of rotation (00:00 - 00:02). This could be because there were more than two key-frames for the rotation of the tire. This problem can be overcome easily by reducing the number of key frames as much as possible.
Particles, Fog and Water
Particles in my scene have made its presence felt to a great extent.
Fog has been used effectively for two reasons...
1. Fog and lighting makes the scene look closer to reality with little effort on the part of the programmer.
2. Fog implies that intricate details are hidden, which means that some error in the animation can easily go unnoticed. Example: water, collision, etc.
I believe the blast effect from the gun are sprites, and gives the right kind of effect. The smoke that comes out at the end of the gunshot makes it look all the more real.
The splashing water looks good, with particles flying off to form ripples far away. The splashing looks good when the dead body is thrown into water, and even more believable when the shot alligator is struggling.
Unfortunately, the water itself in the river / stream looks more like oil. It reminds me of one of the viscous liquid animations that we saw in class. It is so highly reflective, just like everything else in the episodes, and is simply not ready to budge when a Captain falls into it – so not watery. In short, it would have been awesome if the water was a lot less viscous.
When the alligator dies, the blood does not look like another liquid being amidst water. One big red patch - you have to force yourself to believe that it is blood by looking at the color.
The objects, people, vehicles and animal lack weight in the animation. None of them have a feeling of heaviness.
When the fake Captain Scarlet is carrying the dead (real) Captain Scarlet, he doesn't slouch or feel the weight at all. He still continues to spring up and down in spite of the weight on his shoulders. When he reaches the bank of the river, he throws the dead body off like a brick that he was carrying in his hand. This is not intuitive. When doing the motion capture for this sequence, the actor should have carried a sand bag or something similar to get a feel of the weight.
And when the dead body falls into the water, it starts floating right away (this could be expected behavior, because of the properties of the armor); something I doubt would happen in real life (this could be expected behavior because of the high viscosity of the "futuristic water").
Making objects heavy is very difficult in an animation; and hence, it is something that defines a superior quality product.
When objects interact with each other (Example: Touch, Carry, etc), there are two things that can be done:
1. Keep the camera far away or block the view to the point of interaction, so that you don't have to sit and intricately match the edges of the two interacting bodies. This has been done well in a couple of places in the clip. For example, when the car comes in (00:00 - 00:01) there is a bush that blocks the view of the bottom of the tire. So, the tire need not be exactly touching the surface of the ground; it can be a little above or below the ground and it is impossible to notice it.
2. Make sure that the interaction is well set up so that the two objects just touch each other. This also has been done at places in the clip. For example, Zachary's father holding his gun (00:20 - 00:21) has been done with high intricacy. It looks so intuitive and beautiful.
But when you don't implement either of the above two, the result looks displeasing. Even this has been done by the team. When the fake Captain Scarlet opens the passenger-side door of the car, his hand is at least two inches away. This looks so weird. In this example, they could have put a touch sensor on the car and made the fake Captain Scarlet touch it, and the door opens automatically (may be that is what they ARE trying to show); that way, you don't have to run your hand all the way to the top along with the door during motion capture and there will be no synchronization issues. An even better idea would be to open both doors automatically when the car stopped; then you don't even have to touch the door at all.
Small Things that matter
There will always be several small things that would make a big difference when implemented in the scene; two are below:
Interaction with the background
The animators should have given thought for some kind of interaction between or amidst the background objects. The forest, for example, in this clip looks haunted, with no animals or birds except for the alligator.
When the car moves in or out, if some dust woke up, it would look great. It would be too much to ask for interactive grass. However, if the leaves and branches swayed a little, it would have given life to the rest of the scene.
When he walks, the character could sift through a bush or a low branch and make it move / shake a little. This will blend the character with the environment and make the whole scene more realistic and enjoyable.
It is very rare that a car goes through a forest and only one side of it becomes dirty. In this clip, the driver-side of the car has become dirty with mud splashes, et al. But the passenger-side, for some reason, is bright, shiny and untouched. Putting some dirt on this side too would make things consistent.
The product is still rough and can take in a lot of fine tuning, an exercise that takes more and more time and effort as the detail and complexity of the scene increases; but the pains would pay off really very well. Considering the fact that this was done for an early morning kids' show, it is better than good.