WXYZ pixel layout in EXR

Hello.

I’m trying to render a scene in blender in a multilayer EXR render. Flame rejects to read motion vector pass, with next error:

" Cannot access frame1: Failed building MIO format. Non RGB pixel layout has more than 3 non alpha channels: WXYZ"

In blender, I’m remaping vector channels to rgb. And then, I feed the render node. I need it anyway because blender uses an Z-up cordinates and channels had to be swapped for a right interpretation. But, no way. Flame does not want these EXR. Reading metadata in flame, it still showing vector channel as “wxyz”.

Can flame read wxyz pixel layout EXR from other 3D apps? Anyone with experience in blender to render passed for flame?

Take a look here…

1 Like

I always have to convert the vector pass in nuke from blender - I honestly dont know why flame just refuses to read it, you say it breaks even after shuffeling it out using the compositor in blender? I find the compositor kills metadata dor cryptomattes etc so I dont use it.

But I can’t even read the file So, the problem is vector channels have “.X” , “.Y” , ".Z " suffix , and they have to be renamed to “.red”, “.green”, “.blue” , right? … Do I have to use nuke only for that? :sleepy:

Did you try the reshuffle in blender?… I think that thread has an example that works.

This
image
image

Yes, sure. Exactly like you did. I found that info searching about how to fix the stupid orientation-system used in blender. Then, I get the wxyz layout problem with blender’s renders. I thought that separating wxyz channels and combining channels in rgba in blender’s compositor was the fix, but , no way. They still like wxyz layout , and then, flame does not read them.

That was because I posted the question here.

I´ve just tested in nuke, an indeed, reshuffling in nuke to rgba channels, did the trick.

I don’t understand why blender set rgba channels like wxyz after using combine color node , and why, I understood, you can read the wxyz layout.

Bear in mind in the first screenshot the render is coming straight from blender to flame… no nuke intermediate.

1 Like

Not sure if it matters, but at the time I was probably on Blender 3.4, Flame 2023.

blender 3.4, flame 2023.2



Weird. I dont have Flame available now but just did a quick test and upon inspection in RV, vector layer shows as RGBA


Only difference in my Blender setup is that I dont name the default RGBA. I have to leave it blank to be able to have it as default layer in other apps… specially flame

Are you sure you are reading the output of the compositor? Seem like your layer names in Flame are not in line with what you have in Blender:

That is because are screenshots from differentes tests.

I figured out. I didn’t have “compositing” activated in post process option. :sweat: So, I was not rendering throught compositor pipeline. :face_with_open_eyes_and_hand_over_mouth: