3D Preview vs RTP vs Render and Clouds

Started by WAS, March 21, 2020, 08:40:36 pm

Previous topic - Next topic

WAS

March 21, 2020, 08:40:36 pm Last Edit: March 22, 2020, 01:44:40 pm by WAS
I know there's been a lot of talks about hard shadows with the 3D Preview and RTP, but this seems to be more, and it's a continual problem. The 3D Preview and RTP both have direct lighting, absence of clouds altogether in the region, but the render is entirely masked out cloud coverage in the region shadowing the entire terrain.

This is a really huge inconsistency, and it interferes with workflows. The shadows alone were a pretty big issue in workflow, but now there are just clouds where they aren't in the 3D Preview or RTP. Even using HD for 3D preview base for RTP. This is another issue where the RTP isn't helping with cloud generation and scene lighting.

The visible clouds empty space matches in the viewport between render and Preview/RTP.

There are no crop settings in TG set, or provided as an argument for launch.

WAS

Using a z-planar distance shader I get the desired look back (in render), almost as if clouds don't exist off camera to some degree, or severely compromised? The change in RTP/Preview was very minimal if really anything.

pokoy

Yeah, for some reason it's still happening, sometimes more visible, sometimes less, but your example is indeed pretty extreme.
What bothers me most is that there's a noticeable color shift between RTP and final, and it happens even in cases where clouds don't really affect the lighting.

I hope this can be improved.

WAS

IMO opinion the inconsistencies are too extreme for production use. I often just use the 3D preview for an approximation and than just render, but with a more stable system I had been deciding to use the RTP more but it's really just not ready for use IMO. Shadows, lighting, colour grading, is all off by great factors that could lead production in the wrong directions to go to preview render and have wrong results.

pokoy

I wouldn't say it's unusable - quite the contrary for me, RTP has helped me considerably getting much faster turnarounds. The current issues might well just be the corner cuts needed to make it interactive. Of course, I would love to see this improved.

In regards to the color shift between RTP and final renders, it makes it only more obvious that the frame buffer needs color correction abilities, color temperature being the most important one for me personally.

Cloud quality and density 'degradation' could be handled with a 'RTP cloud quality' parameter where users could set their desired level of complexity instead of the black box it now is. That way, we would also know that what we're looking at is not the final quality, and we'd know that there are trade-offs when working with RTP.

WAS

March 31, 2020, 02:17:59 pm #5 Last Edit: March 31, 2020, 02:34:03 pm by WAS
It's more a personal opinion. You being challenged by something and it it helping may not be the same for the next person. And I'm unsure why it has so many problems to begin with. Every RTP I've used are exacting, like I'm Vue, Cinema4D, and Blender. Just blocky. I think the only thing cut in Cinema4D is reflections, and in Blender contrast may be a little darker in shadows. And I know a lot of people in the industry that avoid TG or criticize it for many of it's neglected quirks.

pokoy

I understand the personal view thing, of course. I personally am amazed by the fact that a thing like RTP was possible in the first place, it's probably not the easiest thing to add to a code foundation from a time where no one even imagined this.

I guess it's really the simplifications in sky/atmosphere/volumetric systems that cause this.
If you think about the physical scale TG simulates even a slight simplification might lead to a pretty considerable shift or change in the final result. In a typical 3d scenario this would probably not be noticeable at all but once you have to simulate a ray over hundreds or thousands of miles it becomes visible.
Then again, I'm just speculating, might just as well be a bug or an oversight somewhere.

Matt

I'm aware of the problem and I'll make improvements as soon as I can. The voxel buffers run at lower resolutions in the RTP so that it's more interactive than a test render, but I think you know this already from previous conversations about it. To solve this I may need to bypass the voxels until sufficient data has been cached. I will work on this problem when other TG priorities are out of the way.

RTP is an ongoing development and not everything works the same as a full render. Until it does, I recommend using RTP for the things you know it's good at and use test renders for the things it's not.
Just because milk is white doesn't mean that clouds are made of milk.