When I do a panoramic capture using Scene Capture 2D in Lumen environment, the light is different in each capture, is there any way to fix it?
Different howso? I’m not seeing any obvious lighting discontinuity or error in rendering.
Don’t see the square frames in the photo? Those are the problem.
xd
Square frames? The lines that define the floor and ceiling looks appropriately contiguous, and lighting seems to be flowing from one area to another appropriately. I could of course also be blatantly not seeing something, so if you could circle the area you’re seeing the artifacting that may help.
Oh, I can see it now, I literally didn’t notice somehow. Lumen suffers from discontinuities because of its’ dependence on screen-space passes that don’t necessarily share information with eachother; I believe there are resources to mitigate this in the lumen technical details somewhere?
Should I look at the engine code? Or should I look at the documentation, I haven’t found it yet.
Unfortunately, turns out I’m not really finding a lot of resources on lumen scene capture. The fundamental challenge is that lumen works via screen-space integration, and any time the camera is physically capturing two different parts of a scenes, there will be lighting discrepancies that lumen doesn’t know to fix.
If you’re trying to bake cubemaps or something similar, I would personally just use GPU lightmass. It’s really fast, especially for small scenes, and it doesn’t have discontinuity problems by and large.
Actually, this also happened to me. I solved it by set the “vignette intensity” in “lens” in postprocess setting of scenecapture2d, turn on it and set it to zero.