Unreal Engine Experiments
Ongoing research notes on new procedural and pipeline workflows.
A running notebook of smaller experiments that haven't (yet) turned into a full case study — the kind of research that shapes future pipeline decisions even when it doesn't ship directly.
PCG + Niagara data interop
Testing how far PCG-generated point data can feed directly into Niagara as a data interface, rather than going through an intermediate static mesh or instance buffer. Early results are promising for scatter-driven VFX density (dust, debris, foliage motion) that should visually track the same distribution logic as the static scatter, without maintaining two separate placement systems.
Auditing PCG runtime cost per node type
Building a lightweight profiling overlay that attributes PCG graph execution time per node, rather than per graph. Graph-level timing was hiding the fact that a small number of node types (mostly anything doing a spatial query against dense point sets) accounted for the overwhelming majority of runtime cost. This is shaping which nodes get flagged for bake-time execution instead of runtime.
World Partition + procedural streaming boundaries
Investigating how procedural biome boundaries should be authored relative to World Partition cell boundaries, so that a biome definition doesn't produce visibly different density right at a streaming seam. Still early — current approach pads generation slightly past the cell boundary and clips, which works but isn't the most efficient answer.
More notes as these experiments mature into something concrete.