I’m seeing what I think might be errors in how the reflection plane probe renders in 2.8 beta Eevee. In the 2-frame GIF image (link below), I have a reflection plane probe placed just above a geometry plane with reflective material. Three spheres float above, and a single HDRI world illuminates the scene.
I’m noticing that, with ScreenSpaceReflections OFF, the reflection of the spheres appears “multiplied” or “screened” onto the reflection of the world. I would expect the spheres to appear opaque in the reflection and not show any world through themselves.
When ScreenSpaceReflections are ON, the spheres now appear opaque as expected, with no world showing through them. However, the world reflection has now shifted slightly. This also seems strange – I would expect the world to appear basically the same in either case. Another oddity is that the spheres in this case have ghosts of themselves smeared off to screen right. This may be due to the imperfect nature of SSRs, but here it seems like the approximation would be a lot closer.
If others feel this isn’t correct behavior, I will submit a bug. Please comment if you have any thoughts. Thank you.
Alright, the transparent spheres: Your material probably has a roughness value between 0 and 0.1. Planar reflections are always sharp, and not affected by roughness unless you use Screen space reflections, so that’s the fallback. As roughness increases it gets blended with the cubemap, and above 0.1 you can’t see it.
About the shifting, it only happens to the reflection of the sky, and the edge where the geometry meets the sky, but the geometry itself seems properly reflected. Maybe this could be reported as a bug? I reckon that since with SSR off you’re using the precalculated sky cubemap, and with SSR+planar reflections the probe reflects the actual sky from the mirror position, not the cubemap, it’s hard to align them.
And the ghosting, it doesn’t happen to me. There’s some edge fading that you can turn to 0 in the screen space settings, but I have not been able to replicate that problem. This might be a bug specific to you and could be reported as well, with your gpu and driver info, unless it’s another thing
Thank you! I would have sworn my roughness was set to zero, but it was notched up slightly (0.005). I’ve been experimenting and trying to solve the problem, and must have left it non-zero. Thanks, again.
I’ll try to narrow down the smearing artifact and perhaps post a bug report.
