It's a bug. If you have several translation (usually including tracking) nodes one after the other (for concatenation), 4.0 seems to have a bug where this now causes the tracking info to break. Simply insert another node (i.e. a max node with no second input) to break the concatenation and it should work.
quote name='thelastguest' date='02/02/06, 08:42 AM' post='227118']
This is one is driving me nuts; if anyone has answers I would really appreciate it....
To fix edge fog on a piece of handheld footage, I have a complex multi-layered composition-- lots of stablizer nodes, plates, rotoscoping etc. Footage is 1080p, 23.98, 254 frames long, rendering it as a Quicktime with the blackmagic 10bit RGB codec.
My system is a dual G5 2.5, 7 gigs RAM, Shake 4.
My problem is that when I render this at full base resolution, near the end of the clip, one of my plates begins to drift (or actually fails to drift to follow the footage). It seems "lost" for about 6 frames--the panX parameter doesn't seem to be doing anything. When I step through this in the viewer, it is fine and I check the values of all the pan they are okay, too, and when I render it at %50 rez or lower it's okay. The plate is programmed to drift using a move2d node that pans by importing data from a tracker node:
panX = MyTrack.track1X - (MyTrack.track1X@@0001)
Am I missing a step? Is there something wrong with this formula? I have been using the formula successfully for other track/powerwindow type operations. Could this be a memory thing? My cache.cacheMemory is set to 768megs. The are no cloned nodes in the project. I have tried checking it out higher in the node tree and I get the same problem. My keyMixes are set to clip off the main footage, not the plate. I've also tried just rendering 20 frames of the portion that drifts; same result.
Thanks,
insight would be much appreciated.
[/quote]