Hey thanks for answering my question.
While proposed solution would be good, it does not solve all problems.
For example, moveable objects won’t have baked static lighting, which while is intendent for lighting, would be problematic when using AO for texturing.
Or I would like to have dynamic skylight lighting (because of moving objects) and dynamic directional light with dynamic indirect lighting (either LPV or VXGI), still require indirect lighting intensity to be above 0.
It would be nice to have ability, to bake ambient occlusion, for fully dynamic lighting, but to actually never show it on meshes, and just use it for texturing. It would just ignore all light sources expect skylight, and would never render in game.
The DiffuseBoost 0 will remove indirect lighting from lightmaps - so they should be black. bGenerateAmbientOcclusionMaterialMask tells Lightmass you do want the AO accessible in materials. DirectIlluminationOcclusionFraction / IndirectIlluminationOcclusionFraction 0 says ‘dont apply the computed AO to lighting’
I haven’t tried this, but should do what you are describing, without messing with the settings that dynamic GI methods use (IndirectLightingScale on the light)
Will we ever see a way to fully occlude a dynamic skylight? Our concrete use-case is a character entering a dark cave from a fully lit open world. Due to the omnipresence of the Skylight the cave is bright. But if we turn off the Skylight, looking back outside will have the outside appear dark too.
You could try having the cave and the outside world being in different lighting channels. Having 3 channels could also allow you to have a separate lighting channel in the middle to handle the transition at the beginning of the cave.
Honestly the main holdup was just finding the time to implement it. There are a million rendering features needed to make high quality 3d graphics, and it takes a long time to carefully craft each one such that it is performant, and meets artist’s needs. Also note that we are just now getting Hair and Skin shading models in UE4! Obviously important things to have, but good features take time.
Yes, due to the tiled deferred implementation using a compute shader. A SM4 (DX10) implementation could be made, it wouldn’t be as efficient though.
There’s no similarity, a light portal simply tells Lightmass where to look for light entering, it does not change the amount of light energy in the scene.
That’s a really difficult problem because full occlusion of a sky light means the occlusion has to be global - you have to handle casters at any distance. The only reason we have any realtime occlusion right now (Distance Field AO, SSAO) is because of limiting it to local occlusion. The only way to do something like this is to render N shadowmaps from different angles. It tends not to look like sky occlusion though, you see distinct shadows under everything.
However Stationary Skylight shadows can be done correctly with full occlusion from any distance because they are offline.
Skylights don’t respect lighting channels at the moment, although we might change this. There’s some extra GPU cost to supporting it.
It’s something I personally want to add, but it’s not the team’s focus so yes, it would have to be an Epic Friday project or something like that. I already have a long list of side projects I want to do =( Lightmass Portals were done as a side project.
I think they can work well for cars, but there might be problems with the current implementation:
Capsule shadows only supported on skeletal meshes atm. There’s no technical reason for this, just haven’t gotten around to supporting it on static meshes. In The Last of Us, random dynamic props like ladders had soft capsule shadows so it would clearly be useful.
When capsule shadows get so soft that they become ambient occlusion, there’s an artifact in the shadowing that causes a hard line.
You can think of Capsule shadows as advanced blob shadows. You can use just one capsule, or you can use many and get better self shadowing accuracy.
