I would like to know if there’s a recommended workflow for rendering passes like depth and world position. From what I understand, the standard method has always been adding post-process materials under the corresponding renderer in MRG/MRQ. However, based on my tests neither Pathtracer, nor Deferred generated the correct results.
[Image Removed]*
Beauty render for comparison
In the case of Deferred renderer, the passes came anti-aliased which is not correct and the only way I found to mitigate that was to set both temporal and spatial samples to 1 which in turn compromises the quality of the beauty pass.
[Image Removed]
Samples > 1 vs Samples = 1 (correct)
In the case of Pathtracer, the aliasing was correct even with higher samples. Still, the samples had to be set to 1, otherwise they would accumulate over time and create an “inflated” look. Additionally, we had to set the cvar r.PathTracing.FilterWidth to 0. Again, both of these changes affect the look of the beauty pass.
[Image Removed]
Samples > 1 vs Samples = 1 vs Samples = 1 + cvar (correct)
We eventually decided to render the depth and world position passes separately in a new job which makes me wonder whether the current default setup in unreal is correct.
I have been unable to replicate the outcomes in the provided descriptions.
Are you able to please provide detailed steps to reproduce the outcome you have described and/or a minimal project with the settings and sequence that produces the described results.
Some things of note :
Post processing volume settings will affect the final results e.g. ensure to set motion blur to 0 on the post processing volume as well as lower temporal and spatial samples when attempting to disable anti aliasing.
Individual lights and lighting settings in the post processing volume can have their own sample counts
below is an example project where I have set up a simple scene with all the relevant assets in the root folder. If you load the EditorMoviePipelineQueue in MRQ and render, you should get the results I mentioned. (or open the preexisting ones in …\aovTest_archive\Saved\MovieRenders)
Thank you for providing the project files. If I understand correctly you are trying to determine movie renderer settings that will result in an output the same as the editor viewport, particularly with regard to depth.
Some things that may be of note:
The best Non Anti-Aliased results may be achieved by keeping anti aliasing on, dropping the spatial and temporal sample counts to 1 and overriding the anti aliasing method to None.
The GameOverrides Settings section may provide additional tools to achieve the results that you are looking for ( this can be useful for disabling editor settings that may be on by default ).
Output resolution also plays a role in the perceived clarity of the results.
Ensure that the preview from the CineCameraActor looks as you expect as it is what is used for rendering and has its own cinematic effects that may need managed.
Adding warm up frames may allow the editor additional time to adjust lighting etc before the render
The following image was achieved with the following settings.
Apologies for the misunderstanding, are you able to please expand on the steps you are taking and any additional software you are using during your exr workflow that produced the images attached to the original post. The output files from MRQ in the provided sample project are not matching the reported results on my end.
Unfortunately issues involving third party software fall outside of the scope of support we can provide on EPS. If you are able to reproduce the issue using only Unreal tools I am happy to continue the investigation.
Thanks for your reply. I believe there may still be some misunderstanding of my question. I was only describing one of the possible ways of previewing the render pass. You can still see the same results directly in unreal if you apply the same Post Process Material from the MRG to the camera.
Could you please escalate this to someone from Epic Games support directly? I think my case might need more specialized assistance.
Sorry for the delayed response. You are correct in that this is a current limitation of the render process. We are unable to selectively apply antialiasing per buffer layer in the same movie render job.