Shuffle

Shuffle works on spatial pieces of one raster. It is not a channel-routing Shuffle like the similarly named tool in other compositors: the image is divided into a grid and parts of that grid are repositioned.

Blackmagic places Shuffle in the Resolve 21 Krokodove Image Warp Tools and describes the operation as repositioning a subdivided frame within the image. That is the documented contract used by this article.

The historical Krokodove Node Reference independently confirms the two Type choices, Slide and Swap. Public material does not document the exact internal routing algorithm, so observable behavior should be tested on a numbered grid rather than invented.

Random Seed chooses a repeatable pseudo-random arrangement. Keep every other control fixed when changing Seed so the layout change is not confused with a new grid or a different effect amount.

For templates, store Random Seed together with Type, subdivisions and Amount. A fixed seed makes the layout reproducible instead of leaving the composition dependent on an implicit random state.

Type: Slide is one of the two confirmed modes. Its name suggests segment sliding, but the exact cell-routing rules are not published in the brief references, so judge it by a live test pattern.

Type: Swap is the second confirmed mode. The cleanest comparison is to switch only Type while Seed, grid and Amount remain identical.

Horizontal Subdivisions sets grid density in the horizontal direction. Small values produce large readable blocks; larger values create narrower spatial pieces and a denser disturbance.

Vertical Subdivisions independently sets the vertical density. Different horizontal and vertical counts are useful for building bands rather than always using square tiles.

The effective tile size is therefore a combination of both subdivision controls. Choose the grid scale before animation so the effect matches the scale of faces, titles, logos or background texture.

Amount is the main captured Shuffle strength control. The saved screenshot shows 5.0, which is evidence of that Inspector state rather than a universal preset or proven factory default.

Diagnose Amount only after Seed, Type and grid are locked. Otherwise a stronger effect and a different arrangement become impossible to separate visually.

Render is a separate page from Controls. Treat that separation literally: Controls define the shuffle structure, while Render handles resampling and boundary behavior.

Warp is shown as 1.0 in the captured Render state. The brief public guide does not publish its formula for Shuffle, so do not describe it as a proven second Amount control.

Under-Sample Resolution is shown as 1. Its internal Krokodove sampling algorithm is not explained in the brief public material, so evaluate it at 100% Viewer scale and in playback.

Filtering exposes three captured choices: None, Pixel and Area. These are resampling decisions after spatial reordering; they do not change Seed, Type or subdivision geometry.

Area is selected in the saved Render capture. It is a useful confirmed baseline for ordinary footage and graphics, not a rule that it is always the best mode.

Edges offers Black, Wrap and Duplicate. These modes matter when transformed sampling reaches outside the source raster: Black reveals the gap, Wrap samples the opposite edge, and Duplicate extends boundary pixels.

Wrap is most natural on seamless procedural material. On recognizable footage it can expose the opposite frame edge. Duplicate can hide a small gap but becomes an obvious stretched band under a large excursion.

Use a source with recognizable spatial structure when learning Shuffle: a numbered test card, large Text+, UI grid or clearly separated image regions. Fine noise makes it hard to see what was actually repositioned.

For glitch transitions, readable timing is usually more important than maximum complexity. A short burst of strong shuffling often reads better than a long continuous disturbance.

If the effect needs to return to the untouched source, verify the actual neutral behavior of Amount in the target build instead of assuming a numeric endpoint from the parameter name alone.

Unlike Shear, Relative Transform and Offset, Blackmagic's brief Shuffle description does not call out a second image input. This reviewed article therefore uses one confirmed Image Input and does not invent an Effect Mask.

That conservative topology is not a claim that no hidden or dynamic sockets can exist in every build. Exact socket labels and any additional inputs should still be verified live before publishing a macro or preset.

Production QC should include 100% inspection of tile edges, playback for temporal popping, frame-boundary checks for Edges modes and a bypass comparison against the original composition.

For transparent graphics, also inspect Alpha. Spatial reordering and resampling can reveal seams in alpha even when RGB looks acceptable over a dark background.