One of the most crucial aspects of creating an immersive scene is lighting, until now our Scene system had really flat lighting and some things would not show up as you were editing.
Lighting in the editor should now be much improved. Cubemaps and indirect lighting are generated in the Scene viewport as you work, so what you see is what you get.
You shouldn't have to compile your map to see good lighting; you should be able to drop into a scene and see something nice right away.
The goal should be to make the editing experience as seamless and interactive as possible, yet not heavy on your system's resources.
You can choose to have this calculated when the probe is dirty or even as a fully dynamic lighting solution, there's a test scene that showcases dynamic indirect lighting.
You can choose to have this calculated when the probe is dirty or even as a fully dynamic lighting solution, there's a test scene that showcases dynamic indirect lighting.
This whole indirect lighting stuff applies for Volumetric Fog as well:
Future
This solution currently works fine but doesn't have the granularity for fine details that VRAD3 or a full global illumination solution would give, that'd be the next iteration.
A lot of people have concerns for performance or that quality would not be as good as a baked lighting solution, being interactive doesn't mean we can't choose to precompute lighting, just that it shouldn't be a blocking step, and that the artist has full control if they want it dynamic.
I'd avoid traditional UV lightmaps in favor of a solution that's unified between objects, geometry and volumetrics and doesn't need artist input. There's been some good research done on DDGI and sparse voxel textures that strike a fine balance between performance, memory usage, and fidelity, which I'm exploring next.
A lot of people have concerns for performance or that quality would not be as good as a baked lighting solution, being interactive doesn't mean we can't choose to precompute lighting, just that it shouldn't be a blocking step, and that the artist has full control if they want it dynamic.
I'd avoid traditional UV lightmaps in favor of a solution that's unified between objects, geometry and volumetrics and doesn't need artist input. There's been some good research done on DDGI and sparse voxel textures that strike a fine balance between performance, memory usage, and fidelity, which I'm exploring next.
