Points Field

Given a Points stream, generate three fields that describe the local neighborhood: distance to nearest point, direction to nearest point, and the color of the nearest point. Point-cloud SDF + attraction vectors + color-from-source.

Category: Fields Menu path: Fields > Points Field

Ports

Port	Type	Direction	Description
in	points	input	Source point cloud (also accepts shape — shape vertices are used)
scalarField	scalarField	output	Normalized distance to the nearest point (0 at a point, 1 at radius or beyond)
vectorField	vectorField	output	Unit direction toward the nearest point. Zero at coincident samples
colorField	colorField	output	RGBA from the nearest point's color attribute (white fallback)
density	scalarField	output	How crowded a position is: points within radius, divided by densityFull, clamped to 0–1
flow	vectorField	output	The local average velocity of the cloud, in px per second, scaled by flowRange. Zero where the stream carries no velocity

Parameters

Param	Type	Default	Description
radius	scalar	100	The neighbourhood size. scalarField reaches 1.0 at this distance from the nearest point, and it is also the radius density and flow survey
densityFull (UI: "Full At")	scalar	8	Neighbour count that reads as density 1.0. Lower it to saturate sooner
flowRange	scalar	500	Speed in px per second that reads as flow magnitude 1.0, same convention as Point Attributes' Speed range

How It Works

For every sample position, Points Field runs a nearest-neighbor scan across the source points and returns three derived quantities: the normalized distance, the unit direction to the winner, and the winner's color attribute. A single evaluator is shared across all three outputs via Arc, so the per-sample cost is one scan regardless of how many of the three fields a consumer reads.

All four queries run against a uniform neighbour grid built once per evaluation, so a sample costs the points near it rather than the whole cloud. Nearest-neighbour uses an expanding radius, which keeps it exact for a sample far outside the cloud.

The vectorField points toward the nearest point — useful for attraction/snap behaviours. Invert via FieldMath Multiply by -1 for repulsion.

Empty inputs return sensible defaults: distance 1.0, direction zero vector, color white.

Usage Examples

Basic: Point-cloud SDF visualisation

Grid → PointsField → DrawField (scalar mode). You'll see dark disks at each grid point fading to 1.0 outside radius.

Colour particles by how crowded they are

Solver Out → PointsField (the particles of a particle system), then PointsField.density → PointAttributes (field, target Color) with a ramp, and the same solver's points into Draw Points. Dense clumps take one end of the ramp and lonely particles the other. Swap the target to Scale for size-by-crowding.

See the flow a cloud is moving in

Solver Out → PointsField, then flow → DrawField to look at it, or into a second population's PointDeform so a separate set of points drifts along with the first. Size flowRange against the speeds you actually have — Point Attributes shows the live range under its Speed row.

Creative: Attract particles to a grid

Grid → PointsField → vectorField → PointAdvect ← (some other Points source). The drifting points get pulled toward the nearest grid vertex.

Creative: Color bleed from nearest

Give the Grid a per-point color via PointAttributes target=Color, then wire PointsField.colorField → DrawField (color mode). Every pixel takes the color of the nearest grid point — a Voronoi-cell colouring.

Creative: Radius-based masks

Wire PointsField.scalarField → Colorize with a ramp that's opaque near 0 and transparent at 1. Produces a mask of points with soft falloff, useful for glow or density overlays.

Tips

• radius doesn't affect direction or color — only the distance field normalization. Change it to control how sharp or soft the falloff looks when visualised.
• To get distance in raw pixels instead of normalized, set radius to 1.0 — the scalar field won't saturate until you're a single pixel away. (But downstream consumers usually expect [0, 1] ranges, so this is rare.)
• Combine with PointsField.vectorField and FieldMath Add against a Noise vectorField for jittery attraction — points drift toward the grid but never quite converge.
• When feeding a shape input, the shape's vertex points are used. Strokes sample vertices, not along the curve — use ResampleShape first if you want along-curve sampling.

Related Nodes

• DistanceField — same fields but for shape outlines (SDF from boundary), not point clouds
• Grid / Line — natural upstream point producers
• PointAdvect — consumes the vectorField to pull particles toward the points
• AttributeToField — scalar-attribute nearest-neighbor lookup from points (related; Points Field is a superset in output kinds, AttributeToField samples arbitrary scalar attributes)