VIDEOMONTAZHERDAVINCI RESOLVE · VIDEO · AI

FUSION · NODE REFERENCE

FrameAverage

Temporal frame averaging · one image input

Averages temporal neighbors to soften random variation or produce deliberate long-shutter-style motion trails without changing clip FPS.

Time & MotionIntermediate

FrameAverage

Overview

What FrameAverage does

FrameAverage [Avg] is the standard Fusion temporal image-averaging node. It blends neighboring frames from one 2D stream for long-shutter-style trails, smoothing random temporal noise and deliberately accumulating motion. It is not an X/Y spatial blur.

The Resolve 21 manual documents a SINGLE orange primary Input and a processed 2D image output. The pre-existing website article incorrectly claimed a FrameAverage Effect Mask connector. No such second connector is documented; localized lessons place the mask on a separate Merge node.

The source Inspector shows only Sample Direction, Missing Frames and Frames. Sample Direction offers Backward, Both and Forward. Do not invent Strength, motion vectors, Blend or a built-in Mask control for this node.

Backward takes past samples, Forward future samples and Both gathers on either side of the current time. This changes where ghosted moving-object positions appear, not the clip frame rate, output frame count or playback speed.

Frames chooses the temporal window. The archived screenshot captures a value of 0, not a universal recommendation. The specific sample indices, whether the current frame is always counted and odd/even allocation are not fully established by the manual text.

Broader sample windows can make still-image noise less visible while producing strong duplicated edges around hands, moving vehicles or fine titles. FrameAverage averages pixels; it is not a motion-compensated temporal denoiser that first aligns objects by optical flow.

At clip boundaries some requested samples do not exist. Duplicate Original fills gaps from an available original frame; Blank Frame leaves them empty. Check both on a known flat-field clip and with motion, not just a still Viewer frame.

The output remains one image for the current composition time. FrameAverage does not add timeline frames or generate new interpolation time positions. For frame-rate conversion and retiming use a purpose-built tool.

To localize the effect, branch MediaIn into FrameAverage.Input and Merge.Background. Connect FrameAverage.Output to Merge.Foreground and Rectangle.Mask to Merge.EffectMask. That is an external compositing mask and does NOT imply FrameAverage has a mask socket.

For temporal-noise comparisons view 100% resolution and hold exposure, clip time and color pipeline steady. For trails inspect a 24–48 frame sequence rather than choosing a single flattering still frame.

At trims, edit boundaries and timeline jumps, request patterns and caching can change preview performance. Do not claim proven real-time throughput or render behavior before testing in a target Resolve 21 installation.

The three controls and single input are grounded in manual text and two original captures. Typed practice graphs validate routing only; they cannot prove actual sampled pixel values, project FPS, exact sample index choices or cache behavior.

When to use it

  • Reduce random noise in a locked-off low-light shot with a short sampling window.

  • Produce stylized long-exposure trails from moving objects.

  • Smooth subtle variation in a still-background source.

  • Build forward-looking pre-echo for music-video graphics.

  • Localize temporal averaging with an external Merge and Rectangle mask.

  • Compare temporal sampling boundary behavior for QC.