Apple Shake 4 - Manual (Page 696)

696 Chapter 24 Keying
Filtering Keys: The Correct Way
The problem with the three examples above is that the keying node, in this case
Keylight, is being made to do too much—by having to simultaneously key, suppress
spill, and composite within the same node, there is no good place to apply a filter.
The solution is to pull a key for purposes of creating a mask, but to use other nodes to
perform the actual compositing. In the following example, the Keylight output
parameter can be set to either “comp” or “on Black.” The image and alpha that’s output
from the Keylight node can then be filtered, and the filtered result composited against
the background using a simple Over node.
In general, it’s good practice to use the Primatte and Keylight background inputs as test
composites only, to be able to see how the key is looking while you’re tuning it. Once
you’re done pulling the key, you can rewire the output image into a composite using
an Over or a KeyMix node. This method of working gives you several advantages:
• You can apply filters and effects to the foreground material.
• You can transform the foreground material.
• You can color correct the foreground material.
Keying DV Video
DV footage, which is compressed with a 5:1 ratio as it’s recorded with 4:1:1 (NTSC) or
4:2:0 (PAL) color sampling, is less than ideal as a format for doing any kind of keying.
This is due to compression artifacts that, while invisible during ordinary playback,
become apparent around the edges of your foreground subject when you start to key.
With a high-quality DV camera and good lighting, it’s possible to pull a reasonable key
using DV clips, but you cannot expect the kind of subtleties around the edges of a
keyed subject that you can get with uncompressed or minimally compressed video
(decent) or film (best). For example, while you may be able to preserve smoke,
reflections, or wisps of hair when keying uncompressed footage, with equivalent DV
footage this probably won’t be possible.