Point Cloud

From a picture to actual geometry

Every pixel already has two of the three coordinates it needs — where it sits across the frame and where it sits down it. The depth channel supplies the third. Put all three together and you no longer have an image with an illusion of depth on it; you have a set of positions in space.

That distinction is what everything downstream rests on. Once the points genuinely exist in 3D, a camera and a depth buffer do the rest, and points occlude each other because they are actually in front of one another — not because a mask says so. Move the camera and the occlusion changes correctly. Push a point back and it goes behind its neighbours.

Density, size and the shape of a point

Density decides how many pixels survive as points. Wound up, almost every one does and the cloud reads as the photograph reconstituted; wound down, the frame thins until the individual points are the subject and the picture is only what they happen to describe. Point Size is the other half of that trade — small points in a dense cloud read as grain, large points in a sparse one read as deliberate marks. Both are continuous rather than stepped, and the settings worth having are usually between the obvious ones.

The shapes then set the material vocabulary before any material is chosen. Square and Round are flat billboards — the cheapest and the most graphic. Soft feathers at the edges so the cloud reads as a volume rather than a mosaic. Sphere shades each point as an actual ball, which is what makes a coarse cloud read as objects in a room. Digits renders points as characters, and Fused lets neighbours merge into continuous surface.

Six fields, and why Curl Flow and Field Channel behave differently

The displacement fields push the points around once they exist. Turbulence, Wind, Vortex, Explode and Gravity do what their names promise, and Range scales how far any of them reaches.

Curl Flow is the one worth understanding. It is built from the curl of a noise field, which makes it divergence-free — the flow can swirl and fold as elaborately as it likes but it can never pile points up in one place or empty another out. That is why it stays coherent under heavy displacement where ordinary noise displacement tears the cloud into clumps and holes. It is the same property that makes curl noise the standard choice for smoke and fluid motion.

Field Channel is the one that could only exist here, and it reads whatever the app’s shared field currently holds. Usually that is the depth-tangent flow baked from the photograph’s own depth, so the motion runs across the shape of the scene instead of ignoring it — points slide along a face rather than through it.

But the field is a channel, not a fixed source. Put a simulation above the point cloud in the stack — a Physarum layer, growing its slime-mould network out of the same picture — and it publishes the flow along its own veins into that channel. The points then travel the network the mould grew. One effect's output becomes another's motion, and neither had to know about the other.

Sorting, material and environment

Sorting decides which points get drawn on top of which when they compete, and Sort Axis can run vertically, horizontally or through depth, keyed to either the bright or the dark end of the tonal range. Depth sorting with bright tones forward is the setting that most reads as light in a volume.

The four materials — Clay, Pearl, Brushed Metal, Liquid Metal — change how each point responds to the light, and the three environments decide what light there is: a neutral Studio, an image-based environment lit by the picture itself, or a Horizon. Image is the interesting one, because it closes the loop: the photograph is both the geometry and the light falling on it.

When you’d reach for it

When you want the photograph to become an object you can look at from somewhere else. It is the most complete expression of what the depth channel is for, and the one where the difference between reading depth and faking it is impossible to miss.

Every option, from the app

Generated from the registry the app publishes, so this list is whatever the current build actually ships — not a copy of it.

Shape

SquareRoundSoftSphereDigitsFused

Size From

UniformToneDepth

Range

1×2×4×6×10×16×

Fill

OffMirror 2×Mirror 3×

Field

TurbulenceCurl FlowVortexWindExplodeGravityField Channel

Sort Axis

VerticalHorizontalDepth

Sort Tones

BrightDark

Material

ClayPearlBrushed MetalLiquid Metal

Environment

StudioImageHorizon

Related effects

• MetaballDepth points scattered into a volumetric density field and raymarched as an isosurface — blobs that genuinely fuse where they meet, with capsule connectors, cast shadows and ambient occlusion.
• Depth of FieldA per-pixel circle of confusion computed from the depth inferred out of your flat photo, gathered through a configurable aperture — polygonal, catadioptric or anamorphic bokeh.
• Depth ParallaxShift the viewpoint in 3D by displacing every pixel according to its depth — near things move further than far ones — with disocclusions filled by stretch or from the background.

Open the app to try it; the free effects work with no signup. Open V3KTR · all the FX · how it works