Or to put it more specifically, how is it possible to have a character turn their heads from front to profile without first drawing separate images of the character’s face in profile and from the front? A lot of cartoons these days use rigged modeles so the animators don’t have to redraw the characters over and over again, but there are parts of this that I don’t understand. I get something as simple as a handwave, where you could rotate a 2d image around an elbow axis and I’ve done that before in After Effects. I also understand how you can do all this in 3D. but I cannot understand for the life of me how a computer can calculate different perspectives with cutouts of flat images.
Example of what I’m talking about: https://www.youtube.com/watch?v=o3Ky_IOfPkQ
In: 2
Animators draw characters in particular start and end poses, each with the desired perspective. Some movements need intermediate poses (e.g. head looking right > head looking straight-on > head looking left).
The computer simply interpolates between these reference poses, which is straightforward because the characters are drawn as vector graphics.
For a particularly complicated movement, 2 or 3 intermediate poses might be required for smooth interpolation.
So while the animators don’t draw individual frames (because you don’t need to draw anything at all. It’s all vector graphics) these poses are like individual frames through which the interpolation transitions as it proceeds from beginning to end.
The animator starts of with two images: front view and side view of the character. But the trick is, every element (eyes, nose, face, hair) is a vector, and all are on separate layers. This allows them to warp the facial features enough that they can essentially create in-between frames of everything from front to 3/4 to side view, with very little work. This allows them to hide the transition from front to side view.(you can see the jump if you go through some parts frame by frame, but to the common observer it’s very well hidden.)
This is still like traditional animation in that these movements are made frame by frame, the only change is that they can now directly reuse pieces of the character, so now it’s more like puppeteering of a paper mannequin. It’s not like 3D animation so you can’t just rotate a 2D image if a character and expect to get a 360° view of the entire character straight away. Although this kind of AI technology is available for human figures, there is currently no AI that is reliably able to do that for cartoon figures. This kind of tech is not developed enough to be used on commercial animation right now.
Latest Answers