Nodes
Enhance Auto Track
Improve a solve with hand tracks, plumb lines and a better lens model.
Takes an existing solve and refines it, using extra evidence the original run didn't
have. Auto Track and optional Manual Tracks in, refined Auto Track out.
Manual Tracks
Manual Track Weight decides how hard hand-placed markers pull compared to automatic features. Default is 8, so they count for eight times as much, which is usually right because you placed them deliberately.
Bridge Missing Frames
On by default. Frames the solve never registered get placeholder poses, interpolated inside a gap and extrapolated at the ends, then corrected by any manual tracks covering them.
For real recovery rather than a smooth guess, hand-track at least three points through the gap and keep tracking into the solved section on both sides, so the recovered frames can be tied to something already trusted.
Plumb Lines
Powerful lens evidence when you have no calibration data.
Pick a Blender mask whose splines trace edges that are straight in the real world, drawn along the curve as it appears in the image. Three or more points per spline. Building edges, door frames, road markings.
Lines Frame is the frame you drew on. Name a mask layer with a frame number, like
36 or frame 120, and that layer pins itself there instead, which lets you constrain
a zoom at several focal lengths. Line Weight controls how hard they pull.
Lens Model
Lens Model picks what to fit: keep the current one, Brown-Conrady for most lenses, Kannala-Brandt for fisheye, or Double Sphere for very wide and action-cam lenses.
Lock Principal Point keeps the optical center fixed, and is recommended.
Robust Loss (px) is the Huber threshold, above which residuals count linearly rather than being trusted.
Focal
Fixed Lens solves one focal length for the whole shot.
Zoom (Variable) solves one per frame, and needs a solve made with Calculate Zoom. Zoom Smoothness ties neighbouring frames together: around 0.5 for real zoom pulls, 10 to 25 when the lens barely moves, otherwise hand-held jitter leaks into the focal length.
Lens Samples takes distortion-versus-focal calibration from grid solves at two or
more zoom positions, written as 18mm -0.212 0.214; 35mm -0.034 0.080. Per-frame
distortion is then interpolated from that curve and locked, which is the physically
correct answer to distortion changing through a zoom.
Confidence Report
On by default. Diagnoses the solve afterwards: parallax health, how planar the scene is, lens coverage, and per-parameter uncertainty. Worth reading before trusting a lens.