AaronBerney Posted October 20, 2023 Share Posted October 20, 2023 (edited) Hi, I have a room with a large internal window in the corner. The space behind is dark, so reflections are prominent, and it's important that the reflections are representative. I am representing the glass using a semi-opaque mesh transparent material. The problem I have is that the box projection is not applied. See the attached for a test case I've created - this has the opacity at 100%, so it's a mirror in both cases - the only different is that I have switched from mesh opaque to mesh transparent. Is this expected behaviour? Is there any way to resolve this without resort to processor hungry solutions? Many thanks Edit to say: it also appears to have disregard for which environment probe the surface is in. If there are other transparent surfaces inside other environment probes then as the angle of view changes it jumps around from one environment probe to another... Edited October 21, 2023 by AaronBerney Further info Link to comment
bmyagkov Posted October 23, 2023 Share Posted October 23, 2023 Hello! On 10/21/2023 at 3:42 AM, AaronBerney said: Is this expected behaviour? Is there any way to resolve this without resort to processor hungry solutions? You're right and this behavior is by design primarily because of the parallax feature which does not work with transparency at all. Unfortunately, there is no immediate workaround that can be implemented. The only solution is to use dynamic reflections by implementing a Planar Reflecting Probe if your window is actually flat. Thanks! Link to comment
AaronBerney Posted February 28 Author Share Posted February 28 The planar reflection probes seem pretty much prohibitive for a VR project, so I've taken matters into my own hands with this: This gives me two things: it gives me absolute control over which cubemap to use for reflections, and also gives me simple parallax (not a full cuboid projection). It works, but I haven't looked into the impact on performance. Of course it looks far more involved as a graph than as just '(V / V.N - 2 N) d + O - P'. I have to create a child material for each window that I want to treat this way - and copy in the probe centre position, but I like the level of control I get. The fact that this is possible makes me curious as to why parallax is not supported for transparent materials... Link to comment
Recommended Posts