Point Cloud 3D Node

A point cloud holds the features tracked by software (SynthEyes, PFTrack, or Fusion's built-in Camera Tracker) to rebuild the motion of a real camera. Each point is a 3D marker for a feature identified in the image.

Point Cloud 3D can import a cloud from a file or be fed automatically by the Camera Tracker. Once in the scene, it is used to:

• Identify a ground plane - the points that land on the ground let you align a primitive with the filmed scene.
• Orient 3D elements (added 3D objects) correctly relative to what was tracked.
• Publish individual points as exposed controls, so another node can follow their position exactly.

• Effects Library: Tools > 3D > Point Cloud 3D
• Quick add: right-click in the Node Editor → Add Tool > 3D > Point Cloud 3D
• Often created automatically by the Camera Tracker after a solve - no need to add it manually in that workflow.

• SceneInput (orange) - optional. Lets you combine the cloud with another 3D scene (camera, existing objects).

Visual appearance

Style (default: Crosshair)

Displays the cloud as crosshairs (small crosses, easier to read for analysis) or points (to visualize density).

Lock X/Y/Z

Locks the three size axes together. Uncheck it to get independent Size X, Y, Z.

Size X / Y / Z

Crosshair dimensions. Increase them if the points are too small to see in your scene (depends on the scene scale).

Color

Crosshair color in the viewer. A high-contrast color (yellow, magenta) is recommended to tell them apart from the filmed plate.

Data management

Density (0.0 - 1.0, default: 1.0)

Probability of displaying each point. 1.0 = all points. 0.2 = about one point in five. Useful to lighten the display when the cloud holds thousands of points.

Import Point Cloud

Button that opens a dialog to import an external file (see formats below).

Make Renderable

Controls whether the cloud is visible in the OpenGL viewer and in the final render. Turn it off to hide the points once the track is validated.

Unseen by Camera

Hides the cloud from the Renderer 3D without removing it from the viewer - you still see it for debugging, but it doesn't appear in the render.

Formats supported for import through the Import Point Cloud button:

• .ma - Maya
• .ase - 3DS Max
• .lws - LightWave Scene
• .xsi - Softimage XSI

These formats store not only the 3D positions of the points but also their names - handy to identify specific tracked features ("corner_01", "window_left", …).

The most common workflow: the Camera Tracker solves a scene, then automatically generates a Point Cloud 3D with every tracked point, along with a Camera 3D animated with the solved motion.

Typical pipeline:

1. MediaIn (filmed plate) → Camera Tracker
2. Internal solve in the Camera Tracker (feature detection, matching, camera solve)
3. Export → generates 3 nodes: Point Cloud 3D, Camera 3D, Renderer 3D
4. The Point Cloud is used to visually validate the solve (the points must stay still in 3D space when you play the timeline).

While working with a Point Cloud in a 3D viewer:

• Shift + F - Find: opens a search field to find a point by name.
• F2 - Rename: renames the selected point.
• Shift + P - Publish: publishes the selected point as an exposed control, available from other nodes through Connect To.
• Shift + C - Create point: creates a new point at the cursor position (to manually add a marker to the cloud).

• Validating a tracking solve - play the timeline: the points must stay perfectly still in 3D (if they "slide", the solve is bad).
• Aligning a ground plane - identify the points on the ground, publish two or three of them, then create a Shape 3D plane positioned and oriented from those published points.
• Attaching a 3D object to a feature - publish a specific point of the cloud (e.g. the corner of a filmed table), connect the translation of a Cube 3D to that point → the cube sticks perfectly to the corner.
• Importing from an external tracker - receive a .ma from SynthEyes, import it into a Point Cloud 3D, then use its points as a reference.
• Debugging a large 3D scene - lower Density to 0.1 to lighten the viewer when the cloud has more than 10,000 points.

• Important limitation - the software renderer can not render visible crosshairs. Use a Renderer 3D in OpenGL mode to see them in the final output.
• Turn off Make Renderable once the track is validated and the references are extracted - you usually don't need the cloud in the final render.
• Name your important points (F2) before publishing them - otherwise you end up with Point001, Point002… that all look the same.
• For a precise ground plane alignment, publish 3 points and use the Set From Selection option available in the Camera Tracker (if that is where you started).
• The crosshair color and size are not stored in the imported file - they are a Fusion-side display setting you configure yourself.
• If the cloud's scale looks absurd (points at 1000× the size of the scene or the other way around), check the units of the source software and adjust with a Transform 3D downstream.