Blender 2.8 Cycles Feature Request: DeepEXR
Blender 2.8 Cycles Feature Request: DeepEXR
- DeepEXR
Deep Compositing allows you to merge or holdout CG renders or 2d elem on geo from different software or renderers.
Production examples:
Nuke | Deep Compositing In Rise Of The Planet Of The Apes youtube.com/watch
Atomic Fiction And Deep Compositing On Blade Runner 2049 youtube.com/watch
14 Comments
Having Deep support would be a huge improvement, and a good step toward having professional studios adopt Blender
Yes, this should improve lots of aspects of 2d compositing, like aliasing and occlusion, as well as better post processed volumetrics. The current Z masking node is passable at best
@Wo! 262: isnt cryptomatte a better solutiuon?
@Luciano Muñoz: Crypyomatte has a different usage, it acts like an ID mattes that allows you to select different part of the cg render for compositing treatment.
Where deep data allows you merge 2 cg renders that might be occluded by each other seamlessly.
For exmaple, in this nuke demo by Weta, they have rendered the cars and creatures without any holdout.
Then deep merge them in Compositing. The benefit is that if the client wants to change the animation of the entire crowds, you only need to re-render the crowds, and the update the shot in compositing. That saves you quite a bit of rendering time to re-render the entire environment and cars again.
Hope that helps :) SW
Btw, for those that don't know how Deep Compositing works as opposed to the current Z channel, each pixel doesn't store one value * channel * pixel, but a list of multiple depths for each pixel, from the camera foward, so one value * channel * depth * pixel. Not every depth step is stored, it's not quite like voxels in that sense, but you can imagine the advantages.
Blender would benefit from this better than most industry tools, because deep data is huge, and the built in compositor could use it and be done with it in memory without the huge disk files.
Deep EXR output is a feature that the big studios require. The compositor in Blender won't be used for shot work at a large vfx shop, so chicken <-> egg, get Deep exr output in there first.
Blender needs people they can manage that together with ColorManagement.
its been 5 years.... meanwhile we see new render engines and even close to realtime ones that can output deep. this is preventing blender from being adopted in major studios.
is deep exr already included in 2.8? or is there a workaround for that?
before this feature the entire compositor should have a rework or optimizations. Adding this now its like try to fly when why are barely walking. I would love to see this tech in blender but don't think its ready yet.
@L. Nahuel Belich:
Deep EXR is a universal output format just like the existing MultiChannel EXR in Blender.
My proposal is to add this output format in the render settings, rather than having all the deep compositing toolset inside blender. It should take less resources to implement.
From a VFX lighting & compositing point of view. I think the majority of heavy comp tasks are mostly being handled in tools like Nuke, Fusion, Natron, After Effects for image sequence or Photoshop, GIMP for still image.
I'd love to have a stronger compositor in Blender in the future, but I think that might take way more time and development resourses than just adding an extra data type in the render output settings.
I think currently the blender compositor is more intended for handling simple comp tasks, shuffling and renaming render passes and layers for final render output.
Thanks :)
@L. Nahuel Belich: In render support is much preferable initially than compositor support. I completely agree with Stephen.
I completely agreed, first Blender should render it, later the compositor gets updated to use it.