Average

DaVinci Resolve 21 integrates the Krokodove toolset directly into Fusion. Blackmagic places Average under Image Pixel Tools and describes it as averaging multiple frames in a sequence to create a frame-blending effect.

The critical distinction from a normal Blur is the processing axis. Blur combines nearby pixels inside one frame; Average combines image values from different temporal samples. Static texture and moving detail therefore behave very differently.

On nearly locked material, temporal averaging can reduce random frame-to-frame noise, grain-like variation or mild flicker. When an object changes position, its different temporal states overlap and become visible as a trail or ghost.

Do not document Average as a motion-compensated denoiser. The current Inspector exposes no dedicated motion-estimation or optical-flow block, while Blackmagic's current Krokodove summary explicitly describes frame blending. If motion must remain sharp, use a different process or align the source first.

The captured current Resolve 21 Inspector shows Main Time = Real Time, with Time Speed and Time Stretch in the same menu. Those are verified current UI labels, but their names alone do not prove the exact sampling formula of each mode.

The historical Krokodove Node Reference used the name Flow Time for the first Main Time entry, alongside Time Speed and Time Stretch. Old tutorials can therefore use a different first label; this article treats the captured Resolve 21 UI as the target-build naming authority.

Number of Frames controls the size of the temporal sample set. More samples generally strengthen averaging and extend or densify motion trails while also requiring more neighboring-frame access.

Preceding Frames controls how much of the temporal window is placed before the main time. Think of it as window geometry rather than a simple effect-strength slider: the same sample count can be distributed around the main time differently.

Frame Step controls spacing between selected temporal samples. Step 1 samples adjacent times densely; larger values spread the samples farther apart and can make individual motion phases read more clearly.

Include Current Frame determines whether the main-time sample contributes to the blend. Turning it off can materially change the visible current phase of a moving object because the result then depends only on other temporal samples.

Non-Existing Frames exposes Black and Ignore in the captured current UI, matching the historical Krokodove reference. Black contributes black for an unavailable temporal position, while Ignore skips that missing sample; source boundaries therefore need dedicated QC.

Darkening near the beginning or end with Black is not necessarily a grading error. Part of the sample window may simply lie outside the available source range. Compare Ignore before trying to compensate with exposure controls.

Cuts, Fusion Clips, retimes and image sequences need extra testing. A temporal processor does not understand editorial intent: if a radically different adjacent source state is available to the tool, it can contaminate the sampled result.

Effect Mask remains a normal spatial mask input. It is useful for limiting where Average contributes, but without a live target-build test it should not be described as changing the internal temporal-sampling domain or the set of frames read.

Generic Settings such as Blend and Process R/G/B/A are useful for dosing or channel-gating an already-computed effect, but they are not temporal-window controls. This article keeps them separate from the six Average-specific Controls.

The captured Real Time / 8 / 0 / 1 / Include Current Frame On / Black state is UI evidence, not a universal preset. Choose the window from shot motion, noise character, clip length and acceptable ghosting.

For technical work, decide first whether the goal is noise/flicker reduction or a deliberate temporal trail. Use the smallest sample window that solves the problem, then expand Number of Frames or Frame Step only when the additional temporal footprint is actually wanted.