Hey hey, Im using the MovieRenderGraph in UE5.5, just trying to render 2 layers.
The issue is the layers are not rendering correctly and Im at a loss as to whats happening.
Below is what the graph looks like:
And here is what the renders look like:
(BGOnly)
details of MovieRenderGraph
with a normal CineCamera the images are rendering better (except for the characters alphas arent complete white), but with the VCam they arent:
CineCamera:
@shauncomly any thoughts on what I could do to better troubleshoot?
Are you rendering with a vcam with the HUD turned off or from a take recorded vcam. The VCam Record?
Okay, yeah, I can kind of repro this. Actually, in the Main branch, the alpha from the spawnable take recorded cam is reversed. So, it’s white where there is nothing and black where there is a mesh. Is that the same for you?
Secondly, are you saying that in a cine camera, the alpha is not white? Is it possible you are holding out any low-density atmosphere in front of the characters?
@comlys even without the VCam, then render layer of the characters is off, its not 100% white, its slightly transparent
this is with using the MovieRenderGraph
yea actually ill check if a fog card is in front of them, ill check in a bit
and yes to the question about the VCam, white where theres nothing
Ok you can try and remove the PPM from the camera then. In talking with them they inverted it for something else. But we need to have another think since its messing with MRQ. Thanks for flagging! Report back how it works out for you.
looks like that PPM was the culprit for the inverted alpha on VCam, attaching an image for others who may run into this:
and for the non-white character alpha issue yup lit was a fog card being heldout! i am putting those fog cards on their own collection and modifier (should not have been lazy and done that from the beginning 
):
However there is still this issue with the sliver of alpha at the top, its in both the VCam and CineCamera renders:
also I tried to used the ObjectID pass, but its flickering. Whats causing that?:
super appreciate your help!
We’ll get the PPM removed on the Take Recorded camera in the 5.6.1 hotfix. Thanks for the flag there!
I am not sure about that line. I haven’t seen that before. Is it there if you turn off TSR? Maybe a post effect is messing with it?
For the Object IDs the pass should be consistent on disk, correct? We re-map the colors to a consistent ID when generating the pass data.
It probably flickers like this in the preview window by modifiers in the graph invalidating hit proxy IDs. Its more annoying to look at in the preview window than an actual problem, so it’s been lower on the priority level to try to find a fix.
the line is still there when turning off TSR, but ill do some other digging, its lower priority issue
though for the cryptomatte/objectID i did some testing and found the only way to get it to show in Davinci was to:
the gif below shows me scrubbing the first 10 frames, where the selected IDs vary on the frames where they are visible
Can you put 10 of those frames up on a GDrive or something and send them to me?
kk I DMed you
Ok, I can reproduce this if I have the Object ID in more than one render layer. I normally just have it in one layer. I was also able to get the 1 pixel white line at the top of the alpha. I got a GPU dump with it and passed it along.
Thanks for flagging all of this! We’ll get it fixed up hopefully in time for the 5.6.1 hotfix.
For the 1 pixel line, there seems to be a bug with the Sensor Height of the render camera. In our internal cases, if we limit the number of decimal places to the hundredthousandth, it’s fine. As soon as you add a millionth place, we get the line. Software development is fun.
copy that! is there anything i can do on my end to negate that line?
current sensor info:
and thankya Shaun, i appreciate you lookin into these things! Ill hold on the Object ID and will test when the hotfix is out
Interesting, I am unable to repro it with that film back. Are you cropping at all? Is there anything else that sticks out on your camera that I could try to add here?
Either way, we have a repro, and Eric is looking into it. There has to be some sort of rounding error somewhere.
And there’s no need to thank me. Thank YOU for reporting these things! It’s next to impossible to catch regressions with all the automated and manual testing we do each release.
