Apple Shake 4 - Manual (Page 821)

Chapter 27 Warping and Morphing Images 821
Parameters
This node displays the following controls in the Parameters tab:
overSampling
The actual number of samples per pixel equals this number squared. For better
antialiasing, increase the number.
xExpr, yExpr
The expression to be placed. See above for examples.
xDelta, yDelta
Sets the maximum distance that any pixel is expected to move, but doesn’t actually
move it. A given pixel in an image may be affected by any pixel with the Delta distance.
This means that Shake must consider a much greater amount of pixels that may
possibly affect the currently rendered pixel. This is bad. However, if you set a Delta
value that is too small, you get errors if your expression tells the pixel to move beyond
that limit. Therefore, you’ll want to do some testing to balance between speed with
errors, or accuracy with drastically slower renders. It is recommended that you start
small and increase the size until the errors disappear.
The Warper and Morpher Nodes
Shake’s shape-based warping nodes, the Warper and Morpher, let you easily create
specific warping effects using shape tools that are very similar to those used by the
RotoShape node. Using these tools, you can deform parts of an image to conform to
shapes you create in the Viewer.
Warper and Morpher Memory Usage
The Warper and Morpher nodes use a lot of memory when processing high-resolution
images—using four image channels of the full image buffer in float space for each
processing thread. As a result, memory usage may become an issue when warping and
morphing large images with multi-threaded processing enabled. In this situation,
virtual memory usage may noticeably slow processing speed when the maximum
available RAM is used.
For example, if you have 2 GB of RAM in your computer, and Shake plus assorted OS
operations use 300 MB, this leaves 1.7 GB of total memory for image processing by the
Warper or Morpher node for any given frame. You can calculate the RAM used for a
frame at a given image size using the following formula:
4 * (image width * image height * 4) * (number of threads)