The new cryptomatte is a very nice workflow upgrade, thanks @JeroenBakker !
As of now it seems that I cannot use one cryptomatte node to combine selections from different passes. In this example I selected: 1 object, 1 material, and 1 asset using the same node. Because the CryptoObject view layer is selected, the matte output is only outputing the matte for the object, without taking the other selected Matte IDs into account.
Does this mean that with one cryptomatte node, we can output a matte with only 1 type of pass at the same time, either object, material, or asset?
Or will one cryptomatte node eventually output a combined matte with all of the selected Matte IDs, regardless of their respective passes?
Hi I haven’t thought about such workflow.
the old node could do this in a way but with limitations when picking. It was never designed for this also. I can see benefits, but not sure how the ux would work without becoming a burden.
Current workflow is clearly communicating the design and normal usage. Any ideas about that?
Cryptomatte uses weighted mattes. Combining different layers can lead to incorrect mattes. Eg adding a selected material on top of an selected object can make weights higher than 1.
I see thanks for the explanation. I also think that it would be useful if possible.
When combining mattes from different passes, using the lighten mix mode would be the correct workflow?
Here we have 2 overlapping cubes, sharing the same material. Thus the object id and material id are overlapping. When using the add mix mode we go beyond one as expected, but when using lighten we stay within the 0 to 1 range and the resulting matte seems to be correct?
This adds a deviation from the cryptomatte standard. I would like to keep the cryptomatte node follow the standard where it is an additive/weighted mask creation workflow. For deviations you can always use a separate setup and combine it in a node group.
Adding your request directly to the cryptomatte node could mean that future changes in the standard would make it harder to implement in Blender. As our node would not work in an additive/weighted manner.
Areas that are troublesome would be transparency and volumetric couverage.
Alright understood thank you for your feedback!