So Cryptomatte workflow had a recent redesign. While previous implementation indeed had some flaws, user had to manually connect inputs and had to tediously select correct object with eyedropper, the biggest flaw until recently was inability to use blenders cryptomatte layers in other compositing app, for example Fusion. Recent update fixed it, and we could easily use blenders rendered cryptomatte passes to get masks of any object directly in compositing apps, without going back to blender and rerendering needed masks from saved cryptomatte passes.
So, after recent redesign of cryptomatte which has been committed now, as quote here says
āTechnically there is some loss of flexibility because the Cryptomatte pass
inputs can no longer be connected to other nodes, but since any compositing
done on them is likely to break the Cryptomatte system anyways, this isnāt
really a concern in practice.ā
So am i right to assume, we wont be able to save cryptomatte passes separately with compositor file output? Itās quite important as in vast majority of cases in production, cryptomatte is used in compositing apps and not in blender compositor.
Edit: I know we probably will be able to save all passes with multylayer exr, but without compositing file output workflow all passes will have to be saved in 32 bit as cryptomatte needs it, so it will allready use much more disk space even not considering that this way even garbage passes will be saved in 32 bit, like denoising albedo and denoising normal which are only needed for denoising and are of no use outside of blender for compositing.
With workflow in picture below i could save all color passes in need in 16 bit, and 2 cryptopasses in 32 bit, and resulting files are 10 mb approximately. While saving all pases in 32 bit for same scene with all denoising passes is 85 mb. Lot of wasted spaceā¦
No, passes sockets has been removed from Cryptomatte node.
But they are still there in Renderlayer node. So, you can still connect them to File Ouput node.
Iām from Italy so sorry for my bad English.
Iām a long time 3dsMax User that recently switch to Blender because at work the facility want to switch to Blender for Productions Tasksā¦Great!
Weāve problem to use a 16bit.exr Cryptomatte, is it possible to make Crypto work at 16bit? Because rendering a video with thousant of frames requires a lot of space at 32bit. With the 16bit Cryptomatte output of Vray all works perfect, and in the compositor program we can see the cryptomatte COLORS, with the output of Blender we can only have Crypto00/01/02 but is not readeable for peak the objects.
So do you need to use the bloody compositor to get cryptomattes into an EXR?
I tried to enable the Object cryptomatte pass, and then just save the āimage as renderā in the Render workspace and OpenEXR as the file format and it doesnāt contain any cryptomatte that I can seeā¦
If you must use the god damn compositor, how do you get all cryptomatte layers into one EXR? (And how do you save that EXR without having to re-render the image? Thereās seemingly literally no buttons in the UIā¦)
No, Itās never been necessary and itās never been the appropriate way to export passes.
Just enable the passes and save a multilayer exr with compositing nodes disabled. Note that exr and multilayer exr are different format options so perhaps thatās why your passes were missing.
Lots of yellow and green on different objects next to each other, instead of blue and red that would separate them more easilyā¦ (cryptomatte layers donāt seem to work for me, but that may be an Affinity issue, so I only get one usable āclownā layer and the rest are black).
Cryptomatte is not a āmatteā in the same way a clown pass/material id/etc is and should not be used as such. The colors are irrelevant. Cryptomatte should be thought of as data, rather than pixels, and that data needs a specific reader. From a casual search, Affinity doesnāt appear to have any support for cryptomattes and so thereās no way to use them there. Youāll need to use render layers and conventional id passes for masking.