3D Point Generator

The 3D Point Generator creates real scene points while you work in the tracked camera view with the footage visible as the camera background. Click a feature on one frame, move in time so the camera has translated, click the same feature again, and MotionMaster 3D triangulates its 3D position. It sits between the trackers and the surface tools because its points can feed the surface builder directly.

Setup

Assign the tracked camera, or leave Camera empty to use the active scene camera. The camera must be the current Camera View and have an enabled Movie Clip background. Choose the lens source: Scene Camera / Undistorted when the background is an undistorted plate and Blender's evaluated camera is authoritative, or Solved Raw Lens for original distorted footage, which uses the selected Heavy result's exact COLMAP model, complete parameter array, and per-frame camera mapping (choose the run with Solved Result). Using Scene Camera mode directly on distorted footage can bias the point.

Placing Points

1. Toggle 3D Point Generator on. Camera and lens settings lock for the session; timeline scrubbing and normal sidebar use still work.
2. Hold Ctrl and press Middle Mouse over a feature inside the camera gate. A cross marks the pending first observation on its source frame; Escape clears only that observation.
3. Move to another frame with real camera translation and Ctrl + Middle Mouse the same feature again. Enable Auto Jump to advance the timeline by Jump Frames automatically after each observation (it steps backwards near the end of the range).
4. After a valid second observation the point is created and the generator stays enabled, so the next Ctrl + Middle Mouse starts another pair. Toggle it off when finished.

Camera Background Fit, Crop, Stretch, offset, scale, rotation, and flips are inverted so the viewport position maps back to the correct footage coordinate, and Solved Raw Lens mode inverts the distortion model before building the ray. The closest point between the two world-space rays is accepted only when the camera baseline is nonzero, the ray angle is above the configured floor, both depths are positive, and reprojection error is below the pixel limit. A rejected second observation keeps the first, so you can try another frame without restarting. Tripod and other rotation-only solves are rejected with an explicit no-baseline diagnostic, because pure rotation cannot triangulate depth.

Outputs and Management

Every accepted point is appended as a loose vertex to the aggregate MM3D Triangulated Points mesh, with no helper Empties. Vertices are visible in Edit Mode and usable for snapping and reconstruction, and adding a point never rebuilds the mesh or removes faces you have drawn between existing points. Each point stores its source and camera-gate coordinates, frames, lens source, ray angle, baseline, ray separation, and reprojection error. Select Points, Clear Points, and Use Points for Surface manage the mesh and hand it to the Point Cloud Surface panel.