I'm trying to overscan a tracked shot basically: I've shot some green screen footage, but the studio was not large enough to get the camera sufficiently far enough away from the talent for the shots we need, so I've got to extend/zoom-out the shot using overscan.
So I'm comping my 4k plate on to a 6k or 8k blank background, then tracking the shot, using Syntheyes with an un-distort/re-distort workflow, and outputting a Blender scene for the CG. All good so far, but obviously a pain to work in 6k or 8k regardless.
So I need to render a 6k or 8k image in blender, but 1k or 2k of the border around the shot is all totally redundant. So how can I 'render region' with the region accurately centred, and define the width and height of the region in pixels? So I'm not wasting time rendering a load of pixels which will be thrown away in the composite anyway.