Positioner

Positioner processes one image branch. It is not a two-image compositor and does not replace Merge: the insert/graphic passes through Positioner while the background plate normally stays on a separate branch and meets the result later in Merge.

DaVinci Resolve 21 integrates the Krokodove toolset directly into Resolve/Fusion. Blackmagic lists Positioner under Image Pixel Tools and describes it as a corner-pinning tool with source and destination controls. That is the reliable boundary: it performs point-based geometric mapping, not automatic tracking or analysis.

Type is one menu control. The captured Resolve Inspector shows One Point, Two Point, Three Point and Four Point, and the legacy Krokodove reference documents the same four values. They are modes of one control rather than four independent switches.

A Source point identifies a landmark on the incoming image; the Destination point with the same number identifies where that landmark should land. Source 1 maps to Destination 1, Source 2 to Destination 2, and so on. Breaking that correspondence order is one of the fastest ways to create a twisted result.

In the captured Four Point Inspector, Source 1=(0,0), Source 2=(1,0), Source 3=(0,1) and Source 4=(1,1), while the Destination values form a smaller rectangle. The capture is strong UI evidence for point order and the normalized full-frame setup. This article deliberately avoids assigning a spoken screen-corner name to (0,0) because the relevant Y convention should be checked in the actual build/context.

One Point uses one correspondence, Two Point uses two, Three Point uses three and Four Point uses four. In practice, additional correspondences impose more geometric constraints. Blackmagic does not document the exact internal solver model for every intermediate mode, so claims such as 'Three Point is exactly an affine solver' should not be presented as a guaranteed contract without live or SDK evidence.

Four Point is the clearest corner-pin use: four insert landmarks are mapped to four landmarks on a planar target. It is useful for manually placing a poster, interface, screen or sign when tracking is unnecessary or already solved elsewhere.

Positioner is not PlanarTracker. It does not analyze a moving surface or generate a track. On a moving shot, Destination points must be animated manually or Positioner should be combined/replaced with an appropriate tracking workflow such as PlanarTracker and PlanarTransform.

The captured Render tab contains Under-Sample Resolution, Filtering choices None/Pixel/Area and Edges choices Black/Wrap/Duplicate. These controls belong to resampling and boundary handling; they do not repair incorrect Source/Destination correspondences.

Effect Mask is a separate optional mask input. It limits where Positioner acts; it is not a second image and not the target plate. In a screen-replacement layout, the plate remains Merge.Background, the insert passes through Positioner, and Positioner.Output normally feeds Merge.Foreground.

Geometry is only the first integration step. After Positioner, inspect alpha, small-text resampling, color, sharpness, motion blur and grain separately. Correct corner placement alone does not make the insert belong to the photographed plate.