Short version

  • Need CG to stick to footage (Camera tracking solves the real camera movement and produces the tracked Blender camera.)
  • Need scene depth or layout reference (Sparse and dense point clouds show reconstructed scene structure for scale, placement, and cleanup decisions.)
  • Need rough surfaces (Generate mesh output from strong dense reconstruction, then clean and align it in Blender.)

Camera tracking solves movement

The tracked camera recreates how the real camera moved. This is what lets CG appear locked to the footage. Sparse points show the features the solve used or reconstructed.

Reconstruction gives spatial reference

Dense point clouds and meshes give you more visible scene structure. They can help with layout, scale, object placement, rough proxy geometry, and understanding where surfaces are in 3D.