Spacing: Unreal vs TUNS

Unreal’s area-weighted sampler balances coverage by triangle area. TUNS also relaxes the spacing between points during the bake. In these captures, TUNS reduced nearest-neighbor spacing variation by 72.2% on the sphere and 64.7% on the skeletal mesh against the area-weighted reference.

Measured spacing consistency

Reference	Unreal variation	TUNS variation	Less variation
Sphere — area-weighted	0.520	0.145	72.2%
Skeletal mesh — area-weighted	0.525	0.185	64.7%
Sphere — ordinary random	0.581	0.145	75.1%
Skeletal mesh — ordinary random	0.852	0.185	78.3%

Variation is the standard deviation of nearest-neighbor distances divided by their mean; lower means more consistent spacing. The percentage is 100 × (1 − TUNS variation / Unreal variation).

Against Unreal’s area-weighted sampler

Left: Unreal area-weighted sampling. Right: TUNS. Both TUNS banks were freshly GPU baked at 65,536 points, 64 Relaxation Iterations and 32 Integration Samples Per Point.

The sphere reference enables CPU and GPU uniform-distribution support. The skeletal reference enables uniform-distribution support on LOD 0, with all 92,178 triangle weights verified. Both use Niagara’s native RandomTriangle function.

Against ordinary random triangle sampling

Ordinary random triangle selection also shows density bias from unequal triangle sizes. Area weighting addresses that bias; spacing relaxation addresses the remaining clusters and gaps.

Camera distances match the supplied framing, so apparent dot size is not the numerical measurement. Percentages come from captured particle positions. For timing comparisons, see Memory and runtime cost.