I know that 5.4 is supposed to have many fixes and improvements to the light probe system, but in the meantime, we can’t seem to get the current system to give us any sort of coherent results. I have a simple grid distribution algorithm that adds light probes through our levels based on the navmesh and a few other parameters. I know it’s not the optimal placement as far as lighting accuracy is concerned, but the real issue is that the calculated tetrahedrons for our levels seem very broken. They network together probes from opposite sides of the level, and there seems to be far too many connections between probes in general. This causes many artifacts on the characters, namely flickering between extreme brightness and extreme darkness.
I’ve posted a couple images showing the poor tetrahedron architecture. Is there any way for us to get this calculating more accurately with the current system? Should we be breaking our level up into smaller light probe groups, confined to a particular area of the level? Any tips would be appreciated.
Thanks
