Shear

The core model is an axis-based 2D deformation, not a normal Transform. Transform moves, rotates or scales a layer as a whole; Shear changes pixel positions relative to an axis so rectangular geometry can become slanted without a full perspective projection.

Blackmagic's DaVinci Resolve 21 New Features Guide gives a short but important contract: Shear applies an angular shear effect, and an additional input can be used to determine shear intensity. That means the second connection must not be documented as a generic Effect Mask by default.

The captured Resolve 21 Controls Inspector confirms the authored control inventory: Shear All, Input Type with Value and Angle, Shear, Shear Along Axis, Center X/Y, Axis Pivot X/Y, Axis Angle and Size Along Axis. This article stays aligned with that captured UI instead of inventing extra controls.

The historical Krokodove Node Reference also confirms two Input Type modes, named By value and By angle there. The current Resolve 21 capture shortens those labels to Value and Angle. The public material does not publish the internal conversion formula between the modes.

Shear is the main geometry amount. Diagnose it on a Grid, checkerboard, rectangular card or Text+ source because straight vertical and horizontal features make axis deformation obvious. Natural footage can hide whether a problem comes from geometry or content.

Value and Angle are two representations for setting the main shear. Do not assume that the same number in both modes produces an equivalent result. Switch the mode, recalibrate visually on the same test pattern, then store the selected mode together with the numeric setting.

Shear All appears as a separate checkbox in the captured Controls page. The brief Blackmagic description does not define its internal branching, so a production-safe workflow compares On and Off with identical geometry and repeats that comparison with the additional intensity input connected.

The additional input is a defining part of the node topology. Blackmagic explicitly ties it to shear intensity. This article calls it Intensity Input as a descriptive teaching label because the brief guide does not publish the exact socket label for the target build; the exact port name and channel interpretation still require live Resolve 21 confirmation.

For diagnosis, feed the second input with a simple grayscale pattern before trying procedural noise. A black/white split or Gradient makes spatial modulation easier to recognize and separates topology testing from artistic texture decisions.

Center X/Y and Axis Pivot X/Y are separate controls in the captured UI and should not be treated as synonyms. Test Center while Pivot is fixed, then reset and test Pivot while Center is fixed.

The captured Center state is 0.5/0.5. That is reliable evidence for the screenshot, not a universal artistic preset. Move Center only after the basic axis behavior is understood so a misplaced origin is not compensated with excessive Shear.

The captured Axis Pivot state is X 0.5 / Y 0.25. Again, this is the state visible in the saved Inspector capture, not a claim about factory defaults. Use it as a reference point for comparison, not as a mandatory value for every comp.

Axis Angle is 90.0 in the capture. It orients the shear axis, but the brief public guide does not define rotation sign convention or an internal coordinate formula. Verify 0, 90 and one diagonal angle on a grid in the actual target build before codifying directional language in presets.

Shear Along Axis is exposed separately from the main Shear control. Its name places it in the axis-geometry layer, but the short public documentation does not define its exact math. Change it in isolation and observe how the deformation moves relative to Center and Pivot.

Size Along Axis is 1.0 in the capture. Treat 1.0 as a diagnostic baseline, establish the angular shear first, then change Size separately so it does not conceal an incorrect Pivot or Center.

The Render page is a separate layer from Controls. The captured state shows Warp 1.0, Under-Sample Resolution 1, Filtering Area and Edges Black. These are captured UI values, not proof of factory defaults or internal renderer equations.

Warp is a distinct render-level control. Because the brief public guide does not describe its formula, do not call it a second Shear Amount. Lock the Controls geometry, A/B only Warp, and document the visible result rather than an assumed multiplier equation.

Under-Sample Resolution is also visible in Render, but the exact Krokodove sampling algorithm is not publicly documented in the brief guide. Evaluate changes at 100% Viewer on thin text, diagonals and grid patterns and then check motion for shimmer.

Filtering exposes the three captured modes None, Pixel and Area. This is the resampling layer, not the axis-geometry layer. Filtering cannot repair an incorrect Center, Pivot or Axis Angle.

Area is selected in the saved Render screenshot and is a useful neutral comparison state for ordinary video and graphics. None and Pixel are useful diagnostics for hard edges, pixel-oriented graphics and intentionally digital looks. Compare them on the same frame and Viewer scale.

Edges exposes Black, Wrap and Duplicate. These modes become visible when the warp samples beyond the source raster: Black exposes the boundary, Wrap reuses the opposite edge, and Duplicate extends boundary pixels. None of them replaces sufficient overscan.

Strong full-frame shear can quickly drive source content outside the raster. If Duplicate turns into long streaks or Wrap creates an obvious repeated seam, reduce the deformation or provide more upstream raster instead of searching for a boundary mode that hides an extreme geometry problem.

For Text+ and UI graphics, inspect Alpha as well as RGB. Geometric resampling can make semi-transparent edges look dirtier when composited over a contrasting Background even if the isolated RGB viewer looks acceptable.

Shear is well suited to slanted lower thirds, kinetic typography, diagonal graphic accents, procedural textures and spatially modulated shear through the additional intensity input. It is not a four-corner perspective matcher; use Corner Positioner or a perspective tool when the task is plane matching.

For animation, inspect start, midpoint and end, then watch real-time playback. A still frame cannot reveal temporal crawl, boundary popping or aliasing on thin diagonals.

A reliable production order is topology, axis geometry, amount/input mode, optional intensity map, render sampling, then boundaries. Changing all of these layers at once makes artifacts much harder to diagnose.