Storytelling and the Camera
A decision as simple as creeping the camera slowly forward can change the whole dramatic feel of a shot. The main limitation you face is the two-dimensionality of After Effects layers, but that's a huge step forward from the bad old days of optical compositing, when it was scarcely possible to move the camera at all.
Nowadays, most directors aren't satisfied with a locked-off camera for effects shots, yet sometimes the decision to move the camera won't be made until the post-production phase. That's no big deal, as long as you don't completely break the rules for what you can get away with.
Specifically, don't worry about planes of motion and parallax for elements that are lost in the background, that are at the edges of the frame, that appear for a few frames only, or that otherwise won't ever be noticed (see Figure 13).
Figure 13 The audience is only subliminally aware of what's going on with that skyline outside the window, prominent though it may appear in this still image. As the camera pans and tracks to the right, the pyramid building should creep out from behind the foreground skyscraper. It doesn't, because the background skyline is a still image; no one notices because the focus is on the foreground character. (Image courtesy of The Orphanage.)
Move the Camera
You may have worked with a 3D camera in other applications, but the After Effects implementation is unique. For example, the After Effects camera contains Transform options that are uniquely different from all other types of layers (see Figure 14), as well as a couple of hidden features.
Figure 14 A camera layer contains no Anchor Point, but includes two sets of rotation data: the Orientation (its basic angle), as well as separate X, Y, and Z rotation values (to avoid problems with complex 3D rotations). The point of interest appears only when the default Orient Towards Point of Interest option is active.
The most common confusion about the After Effects camera stems from the fact that, by default, it includes a point of interest—a point in 3D space at which the camera always points—for auto-orientation. To clarify:
- Auto-orientation via a point of interest is fully optional. You can turn it off (making the camera a free camera) or change it to orient automatically along the path of camera motion. To do so, context-click on the camera and choose Transform > Auto-Orient, or press Ctrl-Alt-O (Mac OS: Cmd-Option-O) to access the menu of settings (see Figure 15).
Figure 15 Many 3D camera tragedies could have been avoided if more users knew about this dialog box. By disabling auto-orientation, you become free to move the camera anywhere without changing its direction.
- To move the camera and its point of interest in sync, don't attempt to match keyframes for the two properties—this is sheer madness! You can parent the camera to a null and translate that instead.
- Orientation works differently depending on whether auto-orientation is on (causing it to revolve around the point of interest) or not (in which case it rotates around its center).
- The auto-oriented camera flips to remain upright when crossing the X/Y plane while orbiting the center; the free camera doesn't. The auto-oriented behavior is helpful for beginners positioning a camera, but not for camera animation—don't let it surprise and frustrate you.
Thus, the default camera in After Effects includes a point of interest that often must be disabled to maintain or control the direction of the camera as it's translated. This won't come up all the time, given that the camera is often put to more modest uses, such as a simple camera push.
Push Versus Zoom
Knowledgeable effects artists understand the huge distinction between a camera push (in which the camera moves closer to the subject) and a zoom, in which the camera stays in place and the lens lengthens. Figures 16a and 16b demonstrate the difference between pushing and zooming a real camera. Zooming changes the actual lens angle, and has more of an effect on the immediate foreground and faraway background framing than a push.
Figure 16a This shot is a push in with a wide angle lens. Notice the placement of the orange ball in the background.
Figure 16b Zooming in from a distance makes the orange ball's apparent scale much greater; it looms larger in this shot than in an image that was shot with a wider lens but close up.
Most of the time, you'll animate a push; zooming in, generally speaking, had its heyday in the era of Sergio Leone. That's a good thing for you, because it's easier to work with a static lens angle. The relationships and perspectives of objects close up don't change with a push, the way they do with a zoom.
Push It Good
Suppose a shot calls for a push in on a 2D composition. Is it worth adding a 3D camera, or can you simply scale the 2D layers?
A scale is too linear to achieve the illusion of moving in Z space with anything but the smallest move; as you draw closer to an object, its rate of scaling must increase logarithmically. Moreover, a 3D camera lets you add eases, stops and starts, a little bit of destabilization—whatever works for the shot.
Camera motion will appear more natural with keyframe eases, which can add the impression that there was a human camera operator behind the lens. You may choose to augment the default eases with a little extra hesitation or irregularity, to lend that feeling of a camera operator's individual personality (see Figure 17).
Figure 17 A simple camera animation can be finessed simply by applying Easy Ease (highlight keyframes and press F9), but why stop there? Lengthening the curve of the first keyframe gives the camera added (realistic) inertia in transitioning from a static position.
A move in or out of a 2D shot can easily look wrong due to the lack of parallax, unless it's subtle or contains few depth cues. Tracking and panning shots, crane-ups, and other more elaborate camera moves will blow the 2.5D gag unless they remain rather minute. When in doubt, rough it in and ask a neutral observer (or supervisor) if it looks believable.
You can get away with more layering with soft, translucent organic shapes such as clouds, fog, smoke, and the like. Staggering these in 3D space, you can fool the eye into seeing 3D volume where there are only planes.
Camera projection (also called camera mapping) is the process of taking a still photo, projecting it onto 3D objects that match the dimensions and placement of objects in the photo, and then moving the camera (typically only along the Z axis), providing the illusion that the photo is fully dimensional. That is, right up until the camera move goes too far, revealing some area of the image that wasn't part of the photograph.
Figure 18 shows a camera projection that ambitiously features two parked military vehicles in the foreground. A dozen separate white solids with masks were created to form a crude 3D model, ready to receive a projected image (see Figure 19). This example shows both the magic of this technique—deriving perspective shifts from a flat, still image—and the associated problems of image tearing when an area of the frame is revealed that had previously been obscured in the source photo. The key to this effect is the setup: How is it that the one "texture" of the image (the photo) sticks to the 3D objects? The fundamental concept is actually relatively simple. Getting it right is a question of managing details, and that part is fairly advanced and not for the faint of heart (which is why mention of a third-party option follows this description).
Figure 18 Progression from the source image through the camera move (left to right). By the final frame, image warping and tearing are evident, but the perspective of the image is essentially correct for the new camera position. The tearing occurs simply because as the camera moves it reveals areas of the image that don't exist in the source.
Figure 19 The rather complicated setup for this effect: From the top and side views, you can see the planes that stand in for the vehicles and orange cone, which appears stretched along the ground plane.
The steps to projecting any still image into 3D space are as follows:
- Begin with an image that can be modeled as a series of planes.
- Create a white solid for each dimensional plane in the image. Enable 3D for each. Under Material Options, change the Accepts Lights option to Off.
- Add a camera named Projection Cam. If you know the Angle of View of your source image, add that value.
- Add a Point light called Projector Light. Set its position to that of Projection Cam, and then parent it to Projection Cam. Set Casts Shadows to On.
- Duplicate the source image, naming this layer Slide. Enable 3D. In Material Options, change Casts Shadows to Only and Light Transmission to 100%.
- Slide not located properly? Add a null object called Slide Repo, set its position to that of Projection Cam, and parent it to Projection Cam. Now parent Slide to it, and adjust its scale downward until the image is cast onto the white planes, as if projected.
- Now comes the painful part: masking, scaling, and repositioning those white solids to build the model, ground plane, and horizon onto which the slide is projected. Toggle on the reference layer and build your model to match that, checking it with the slide every so often.
- If planes that you know to be at perpendicular 90-degree angles don't line up, you need to adjust the Zoom value of the Projection Cam, scaling the model and slide as needed to match the new Zoom value.
- Once everything is lined up, duplicate Projection Cam, and rename the duplicate (the one on the higher layer) Anim Cam. Freely move this camera to take advantage of the new dimensional reality of the scene.