LensDistort

Technical compositing requires the plate, tracks, masks and CG to share consistent geometry. A standard chain is source → LensDistort Undistort → processing/compositing → LensDistort Distort. Both instances use the same profile. The inverse operation does not mean a negative Distortion value; Mode sets the direction.

An insert-only branch is also possible: transparent text or CG in corrected space passes through Distort and is composited over the original plate. This keeps the plate out of an additional resampling pair but requires exact canvas, gate, center and profile matching. Graphics already distorted must not receive a second Distort after the final Merge.

The mathematical model and camera settings form one parameter set. The source screenshots show nine models and aperture dimensions in inches. Film Gate defines the physical frame opening; it does not select a ready-made lens profile. Do not transfer values across model families, fisheye variants or source scales based on similar names.

Edges and Clipping Mode solve different problems. Canvas/Duplicate determine samples outside the input image; Domain/Frame retain or discard pixels moved beyond the frame. Higher Supersampling cannot recover data lost between Undistort and Distort. Inspect the complete chain and postpone final cropping until the return has been checked.

Calibration appears only for Fusion Division Radial or Fusion Radial. Auto works from the current input; Manual allows frame selection. Measuring an actual lens requires a chart photographed with the relevant lens and settings. A digital Grid is useful for testing a graph but cannot measure a lens that never photographed it.

Output Distortion Map changes the output to screen-coordinate data. A visual resemblance to an STMap does not establish channel, normalization or direction compatibility with every consumer. The map practice below stays within output inspection; it does not claim a verified interchange contract or an executed Resolve render.