Hi, I’ve had this issue for a while and it can be minimized/fixed by splitting meshes or being more careful how surfaces are cut and/or welded etc. But I still wanted to see if there is another solution for this issue before doing a lot of extra work.
I have here a super simple example of a small structure where I have one version where I have the whole structure is one static mesh and one version where I split the mesh so that all the beams are in one static mesh and all the base walls/roof are in another static mesh.
I get really different results when I bake lighting (using baked AO will enhance the issue). Both of these use the same lightmap resolution (I think the combined was 128px and the split 64+64px for the two static meshes).
Also there are no overlapping UVs on either of these, it’s just as when some geometry is thin enough the underlying surface’s lighting will be projected onto the thin geometry infront (by thin it seems to be anything from 10cm - 50 cm tbh).
Tried to find any info about this issue but I wasn’t able to unfortunately.
I’ve encountered similar issues with non-combined meshes. It’s not merely the lightmapping. If you look at the doc pages on skylight, there’s a discrepancy in how the lighting appears in a scene with the cubemap being based on the captured scene and with the cubemap being based on a specified cubemap. The two options in skylight can be rather different in lighting a scene, whether the lightmapping is 100% correct or not. The ‘captured scene’ option actually captures a cubic projection of the background AND the active scene (where your structures and meshes are), and reprojects that as lighting. It’s obviously doing some calculations and transforming the data to lighting. However, the ‘specified cubemap’ option uses the data (color information, and differences in light/dark, and more) to light a scene. This can result in blemishes and shadows being placed in areas they’re not supposed to be, depending on which option is selected and what is in the image (captured scene or a picture / cubemap that is not the scene itself but an isolated image that data is being pulled from). What is your skylight setting for cubemap? What is the skylight’s mobility?
Hey thank you for your reply!
The skylight hadn’t even crossed my mind. I was using “Captured scene” as the setting for it, it was also set to be static. I’ve now been playing around with a few of the settings on the skylight and a bunch of the general lightmass settings and I think it’s improved a bunch already! The issue is still there but way less noticeable at least. The problem is reduced a lot when I make sure to keep overlapping faces to a minimum too, it’s just that some of what I’m doing would require a crazy amount of splitting to make sure there are no overlapping surfaces so that just wouldn’t be worth it at this point I think.
Anyhow thank you for your reply, it helped me learn a bit more about the functionality of the skylight too which is great!
Another thing you might now know is in the mesh editor, there’s two lightmap resolution settings under Build > 0 settings. One is Minimum Lightmap Resolution and the other is simply Lightmap Resolution. Both affect how the lightmap is created and the result. In the doc pages, it is advised to always use a Minimum LR that’s lower or equal to the actual LR (which is located lower down in the settings). Try also working with the Destination and Source lightmap channels. Channel 0 is ‘usually’ reserved for texture maps and as a source for lightmaps, but not as the destination for lightmaps.
