1 person found this helpful
Any effect, whether a composite or motion graphic that you have ever seen in a feature film or a commercial can be composited in After Effects. A lot of 3D work can also be composited in AE but complex 3D animations need to be designed and animated in a real 3D app.
There is not one thing in that video that couldn't be done in AE. I've done a lot of hand roto that completely blends in with the background, even roto involving hair blowing in the wind. It just takes more than one layer. Color grading and matching light from the surroundings is key.
I'm working on a series of tutorials on advanced compositing in AE. One of the shots is a UPS semi truck driving past a factory late in the afternoon. The task at hand is to replace the UPS Logo on the side of the truck, on the Trailer and put a new logo on the back door. Trees cast shadows on the truck as it moves from right to left and the sun reflects off the metal side. A passing car throws a reflection onto the truck that goes right over the logo. To make the shot believable every element that I described required at least one additional layer. To make the new logo look like it was part of the trailer I had to use 12 layers just for the logo, 3 to match the reflecting setting sun, 2 for the reflection caused by the sun bouncing off the car's windshield, 2 for the rivets and seams in the trailer that would be visible through the logo (gotta love blend modes) 4 layers for the shadows, and one for the replacement logo. Everything that moved had to either be tracked or rotoscoped or both. A total of 5 pre-comps were required to corner pin track the truck door, the side of the trailer, the passing car, the back door and a telephone pole. The point of all of this explanation is that if you want your composites to look as good as your sample video then you need a bunch of layers.
This is probably one of the simplest shots because the camera movement is very smooth and there are lots of things to track.
I'm guessing a minimum of 2 layers to do the roto of the girl's hand and the stick. 4 or 5 for just the water composite. Maybe another 2 for the reflection of the car, 2 or 3 for the bridge, and probably separate layers for the shadows on the water, the sail boat and it's shadow, and then a bunch of color grading. I don't think the reflection of the car lines up very well. I forgot to count the layers required for the reflection of the girl in the water. It is an impressive shot but it could not be pulled off with two layers and a little keying or roto work.
The way you break something down is to start working on your composite, then look carefully at reference photographs of similar situations and try and pick out all of the details that will help make the shot work. The light, the distance and angle from the camera, all of these things come into play. The bridge works because the angle of the camera to the bridge is nearly identical to the angle of the camera to the puddle and the distance is directly proportional to the scale. That's why the perspective works.
Here's a compositing basic for you. Perspective is controlled by camera position, framing is controlled by focal length. If you are shooting something, like the bridge, that is 2000 feet long and you want to make it 10 feet long in the composite then you take the scale into consideration when you plan the shot of the puddle.
Let's talk a simple composite like a guy eating a bowl of cornflakes that is 4 times life size. You shoot the guy from 4 feet with the camera pointing down 10º. Now you shoot the close up of the bowl from 1 foot at an angle of 10º. Composite the two together and you have a perfect perspective match of a bowl that's 4 times life size and therefore 4 times closer to the camera. It's called forced perspective.
If you are using stock photos or footage then you have to approximate to get it right. It is not only careful compositing but carefully matching camera angle and relative position that makes this sample video work. I would guess that it took months to complete. This kind of work isn't quick.
Thanks for the long respond, i appreciate it.
I've been checking some rotoscoping tutorials and yes, all of them involve working with a minimum of 2-3 layers to get a shot right. I've been thinking that you need at least a layer for the main object or person, and additional ones for the blurry parts, hair, etc; pre-compose it, deform it for shadow and/or reflection, etc. I figured it was a lot of work but the result is worth it. I also discovered that on AE CC you can track masks, but since i use CS6, I'm forced to do it manually.
If you already have one of the tutorials you said you've been working on, please, share it here so me and any other who's interested can keep on experimenting. Thanks again for the reply.