I’d be blown away if they actually animated the camera at all. There would be no need. If you notice, it’s not one giant comp, it’s a bunch of “sections” that pass by the camera and once they aren’t used anymore, I’m sure they are turned off.
This was obviously well planned out, and part of that planning, i believe, is that they different sections and laid them out in 3D space and parented them to a null. Then just animated that null around in front of a static camera. When that section was out of frame is was not in the comp anymore and shut off. It’s like a stylized version of a dad setting up his home camera on a tripod then shuffling his kids in and out of the shot to show the grandparents.
BUT, if there is difficult motion of any sort (especially camera motion), I highly recommend never doing all your motions on just the camera layer. Use layers of nulls to isolate different motions.
For example, say this Brazil piece indeed was the camera moving. I would do it like this:
Layer 1: Null -only animate X and Y position on this null
Layer 2: Null -this is parented to the null above it. This will only be “up and down” motion.
Layer 3: Null -parented to layer 2 – this would be some other motion like maybe cam shake or rotations
Layer 4: Camera – parented to Layer 3. No animation except camera specific parameters like focal length or focus
This way all the motion is isolated and easy to control without conflicting or confusing keyframes.
my 0.2
Chris Smith
https://www.sugarfilmproduction.com