Samplers and Spawners¶

The two ends of every PCG graph: produce points, then turn points into content.

Samplers¶

Node	Source	Use
Surface Sampler	Landscape / mesh surface	Scatter on terrain
Spline Sampler	A `USplineComponent` actor	Place along roads, paths
Volume Sampler	A PCGVolume bounds	Random within a region
Point Sampler	Existing point set	Subsample / resample
Density Sampler	Texture mask	Use a painted mask as density
Attribute Sampler	A data asset table	Drive content by data
Static Mesh Sampler	Mesh's vertices/triangles	Sample on an existing mesh
Difference / Intersection / Union	Other point sets	Boolean composition

Surface Sampler — common pattern¶

Surface Sampler (Landscape)
  → Density = (1 - steepness)        # bias to flat ground
  → Density × Texture Mask (RVT)     # mix with painted exclusion
  → Density Filter (> 0.3)            # cull
  → Static Mesh Spawner              # place trees

Spawners¶

Node	Output
Static Mesh Spawner	Individual `StaticMeshActor`s — flexibility, higher cost
Static Mesh Spawner (Hierarchical)	HISM component — fast, supports LOD per instance
Mesh-to-Points	Points sampled on a mesh — combine for cluster placement
Actor Spawner	Full Blueprint actors with logic
Niagara Spawner	Spawn particles bound to points
Spline Mesh Spawner	Deform a static mesh along a sampled spline

Mesh selection strategies¶

Weighted Random — pick from N meshes by weight
Mesh by Density — bigger meshes at high density
Mesh by Attribute — drive by a species attribute set per point
Choose by Distance from Spline — different meshes near vs far

Density-driven scale + rotation¶

For each point:
  scale = lerp(0.7, 1.2, point.density)
  rotation.Yaw = random(0, 360)
  rotation.Pitch = lerp(0, 5, jitter_attribute)

Performance tips¶

One HISM per (mesh, material LOD set) combination.
Set the Hierarchical spawner's Cull Distance explicitly.
For >100k instances, split your PCG into multiple components / volumes so World Partition can stream cells independently.