Summary: The frame value in input image nodes when set to Sequence fail in all circumstances when frame counts do not start with 0001.
Assuming image-0000, image-0001, image-0002, image-0003, it appears that one frame will be dropped from the series when using the offset value. In the above example, image-0000 will _always_ be missed despite an offset of -1.
Some bad mojo appears at work in the frame counts / offset from what I can tell. It would likely go unnoticed unless one uses large frames of numerical values to help visualize the problem.
Reproduction: Load the following four frames into the compositor. Set Auto-Refresh and Cyclic to on. Set the input node to the following values and note the sequence that results.
The following series occurs:
FRAMES(4) OFFSET(0) : 2, 3, 4, 4, 2, 3, 4, 4, 2, 3, 4, 4, 2
FRAMES(4) OFFSET(-1): 2, 2, 3, 4, 4, 2, 3, 4, 4, 2, 3, 4, 4
FRAMES(3) OFFSET(0) : 2, 3, 4, 2, 3, 4, 2, 3, 4, 2, 3, 4, 2
FRAMES(0) OFFSET(0) : 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
It appears that there is some poor math logic in the frames setting. When there is a single frame in the series, the frames count should be 1 as opposed to 0. The math issue is likely causing further grief when things get more complicated, thereby dropping the first frame.
The bug also applies to bad mojo in background images and other areas.
Version: SVN 31622
Description
Description
Event Timeline
Comment Actions
This is not a bug, but just a matter of how image sequence support works. Files numbered as "0" are not really supported, except for the special case where no sequence is defined (your last test case "frames(0)"). So I strongly suggest you start your numbering from "1" to save yourself from this kind of "trouble".
If you take the above into account then the behavior in all cases is perfectly logical. The duplication of certain frames is a result that you defined the sequence to be 4 frames of length, but only have 3 valid images for the sequence.
Report closed.