We’ve had the odd black square appear on some frames (only 3 of 250) of an animation. It’s only in the beauty pass in the frame buffer, but it saves out with the square embedded. The square moves around but is the same size in all of the frames where it occurs.
The square has appeared when rendered on our render nodes (which are running 1.5 hotfix) and on the workstation (running 1.5)
I did a search and couldn’t find anything else like this. Anything we are doing wrong? (We’ll be updating the workstation to the hotfix as soon as this animation is finished - wanted to match the existing colours.)
I had exactly the same problem once, it was caused by an NaN pipxel and only if denoiser was used. The square should disappear if you disable denoising.
I hoped to be able to reproduce this back then but couldn’t so it wasn’t fixed.
Could you reproduce the error and check, whether the right click pixel inspector reports a NaN? If yes, this is a major issue and has to be found and bug fixed. So many algorithms after rendering depend on the pixels having a real float value. If it becomes suddenly not a number many problems arise.
Could you render the same error and upload a corona .exr dump of it? Stuff like this has fixing priority.
Disabling denoise is not the solution, as denoise works fine and it just fails at not expecting a NaN as input.
I don’t think this has anything to do with the original problem, but it’s not a good idea to use DR with two different versions of Corona (1.5 vs 1.5.1).
NANs are so bad… I always encounter problems when i comp V-ray rendered animations in nuke… Could be cool to detect those NANs pixels and replace them with a float value (something like average color value of the surrounding pixels), in all passes of course.
Yes, either that or add a method to denoiser that would help to handle NaNs that way.
Do you have a global volume material applied by chance? In my case it was an interior render and I’ve used a global volume material. For some reason I couldn’t reproduce it after I submitted a bug report, but it happened almost every time before that. Strange stuff.
Will try and see if the error can be replicated, but it has been rendered again without the square (nothing in the scene or any settings were changed between renders) and only three of the frames (out of 250) had this problem.
If I have time, I will try and reproduce and send scene if I can.
