A simple test bed is constructed using a proxy image made up of a checkerboard and colour wheel, these being graphical node elements built in to Nuke (Fig 1). This is then added to a Card node, which is set at world-space zero.
Fig 1: Test Bed for Camera Animation
A camera is then added along with a Scanline renderer and connected to the Card.
The optical settings on the camera are derived from visual clues in the matte painting (Fig 2), which has a relatively narrow field of view and depth of field extends well in to the landscape with little discernible focus drift, at least as far as the distant hills.
Fig 2: Deriving optical clues from the matte painting.
Capturing this shot in the real world would require a telephoto lens with a focal range of 30 – 70. I went for the lower end of this spectrum to prevent clipping the image too much on the horizontal axis. The shot camera was set to a focal length of 35 and a sensor size of 36 x 24mm (Fig 3).
Fig 3: Optical settings applied to the shot camera
These values are primarily based around the cinematics of the shot and have no direct bearing on the study. However it is worthy of note that the optical attributes, applied to the shot camera, will hereafter need to be applied consistently to any projection camera or Card nodes.
To animate the camera, it is first pulled back on the Z axis and raised in Y to frame the shot. This represents a good position for the animation sequence to end so a keyframe is set at frame 200.
The timeline is then set to frame 1 (the beginning of the sequence) and the camera pushed into the scene. The orthographic (top view and perspective views (Figs 4a and 4b) show the progression of the camera through the 200 frame animation.
Fig 4a: Camera animation seen with 3D viewport set to perspective view
Fig 4b: Camera animation seen with 3D viewport set to orthographic top view
The move is primarily on the Z axis although a slight translation on the X and Y axes, to add a little change in perspective, with a counter-rotation to maintain the shot in frame.
The images (Fig 5) show the progression of the sequence. The scale of the colour wheel shows this is primarily a ‘dolly out’ camera move with the area of focus maintained centrally in frame.
Fig 5: 2D view showing the camera move
The curve editor (Fig 6) is used to manage the keyframe interpolation. I wanted the camera to gain momentum from a standing start, proceed at a consistent pace before decelerating to a halt. Steeper angles indicate faster motion and flatter lines slower motion. The curves provide speed transitions, known as interpolation, and the premise of this is applied across the keyframed position and rotation attributes.
Fig 6: Keyframe interpolation seen in Curve Editor
The final step is to add a Crop node to bring the sequence into 16:9 aspect ratio and output resolution of 1920 x 1080 pixels (Fig 7). This allows us to frame the shot compositionally, based on how it will be seen in the show as opposed to how it will be seen during the 2.5D compositing setup.
Fig 7: Crop node confines the view to 16:9 aspect ratio final output format and resolution.
A test render of the whole sequence (Camera_Setup.mov) provides a feel for the pacing of the camera move and the range of movement. However it is difficult to interpret how this will transfer to the actual artwork. But, by simply swapping the test image for the flat matte paining (Fig 8) we now see the actual artwork at the final output resilution/aspect ratio and from the perspective of the moving shot camera.
Fig 8: Flat matte painting seen from the moving shot camera.
A further render of the whole sequence (Flat_Scene.mov) provides a true sense of how the scene will play out. However this render also provides visual confirmation of why a 2.5D approach is necessary as we see absolutely no shifts in perspective or parallax during the camera move. The shot presents entirely as if the camera is static but the focal length has been altered (zoomed out) to frame the shot whilst the camera was recording.
Figure 9 shows how I set up the basic scene and the optical relationship between the shot camera and the geometry holding the matte painting. This represents the base for the script on which the camera animation was derived.
Fig 9: Annotated recording showing setup of the camera animation script.
Peter Holmes PhD Journal 