Nuke Motion Blur
StyleFrame Team8 min read
Motion blur in Nuke comes from three main sources. Transform-family nodes have built-in motionblur and shutter controls, VectorBlur smears pixels using a motion vector pass, and the NukeX/Nuke Studio MotionBlur node estimates motion from the footage itself. Use the transform controls for animated 2D and 3D elements, VectorBlur for CG renders that include a velocity pass, and MotionBlur or Smart Vectors for plates with no vectors. Whichever you choose, the goal is to match the shutter of the plate you are compositing into.
Pick the Right Method First
Most motion blur problems in Nuke come from using the wrong tool for the source material.
Motion Blur on Transform Nodes
The simplest route is the one most compositors use daily. Transform, TransformMasked, Card, CornerPin2D, Reconcile3D, Tracker, and Stabilize2D all carry their own motion blur controls, so you don't need a separate node.
Three knobs matter:
- motionblur sets the sampling rate over the shutter time. Higher values give smoother results, and a value of 1.0 is often enough. Setting it to 0 turns blur off.
- shutter is how many frames the shutter stays open. A value of 0.5 means half a frame. More shutter means more blur.
- shutteroffset controls when the shutter opens relative to the current frame: centered, start, end, or custom (in frames, with negative values opening the shutter earlier).
Two behaviors are worth remembering. First, either motionblur or shutter at 0 disables the effect, so if blur seems dead, check both. Second, when several of these nodes are chained, the motion blur is concatenated and the last transform in the chain defines the blur applied. That means a stray Transform at the end of a chain can override the blur you set upstream.
This approach samples at many random times across the shutter period rather than averaging a few whole-frame steps like TimeBlur. The result is smoother blur for fewer total computations.
Matching the Shutter of Your Plate
Shutter numbers are only useful if they correspond to the plate. A 180-degree shutter is the common reference. In the MotionBlur node, a shutter time of 0.5 equals a 180-degree mechanical shutter, so at 24 fps the exposure is 1/48 of a second. The same logic applies to the Transform shutter value: 0.5 frames of exposure approximates a 180-degree shutter.
Practical tips:
- Ask for the camera's shutter angle or speed from the data wrangler or the shot metadata, then convert it.
- Check a sharp, fast-moving object in the plate and compare its smear length against your element.
- Keep elements shot at high frame rates in mind. Fast-shutter stock plates (splashes, dust bursts, explosions) look unnaturally crisp when retimed into a normal-speed shot, so they usually need blur added to match.
- Use
shutteroffsetwhen the plate's blur trails the motion rather than centering on it.
VectorBlur for CG Renders
If a 3D package rendered your element, the cleanest workflow is often to render without motion blur and add it in comp. You render an additional motion vector (velocity) pass, then let VectorBlur do the work.
VectorBlur blurs each pixel along a straight line using the u and v channels to define direction. It expects values in pixel-space screen units, where one unit equals one pixel. If the vectors were rendered in a different scale or convention, the blur will be too long, too short, or pointed the wrong way. Verify the pass before trusting it:
- View the u and v channels and confirm they hold signed values rather than 0–1 color.
- Check that the vector resolution matches the beauty pass.
- Adjust the multiplier on VectorBlur and compare against a reference frame.
- Look at occlusion edges, where vector blur can smear background into foreground.
Rendering without blur also gives you flexibility. You can change shutter feel late in the process without re-rendering, which is a major reason to prefer this over baked-in blur for complex FX.
The MotionBlur Node for Footage Without Vectors
In NukeX and Nuke Studio, the MotionBlur node uses the same technology as the blur in Kronos but exposes simpler controls. It estimates motion from your sequence and applies blur, which makes it useful when no vector pass exists.
Key controls:
- Shutter Samples (default 3) sets how many in-between images build each output frame. Raise it for smoother blur, but expect much longer renders.
- Shutter Time (default 0.75) defines the equivalent shutter of the result.
- Method chooses Local or Regularized motion estimation. Local uses block matching and is faster but can produce artifacts. Regularized is semi-global and gives more consistent vectors between regions. Scripts from older versions default to Local for compatibility.
- Use GPU if available moves processing to the GPU. Enable it if you plan to render from the command line with the
--gpuoption.
Inputs matter too. If you've already estimated motion elsewhere, you can supply vectors to FgVecs and BgVecs and skip the estimate. An optional foreground Matte can reduce the pixel dragging that occurs between foreground and background objects.
When Vectors Fail
Optical flow behaves like a 2D tracker across the whole frame. It works well with large, consistently textured features and struggles when the image changes drastically from frame to frame. It's particularly poor with small, similar particles such as water droplets or debris, because it can't tell which particle is which on the next frame. FrameBlend is a cheap alternative, but fast motion leaves ghosting.
Here are practical fallbacks:
- Smart Vectors. These bake motion information into the motion channels, which VectorBlur can then use. Applying them only to the layer that needs them, such as a volume pass, while using the render's own vectors elsewhere is a reasonable way to mix sources.
- Matte the problem. Separate foreground and background (Cryptomatte works well for CG) so vectors don't bleed across edges, and grade or blur each layer independently.
- Paint the blur. Fill the red and green channels with x and y coordinates using an Expression node, smear them with RotoPaint, subtract the smeared result from the original, and feed the difference to VectorBlur as u and v. Only the areas you touched get blurred, which is helpful for sharp edges on FX fire where global blur makes the interior mushy.
- Reveal generated blur selectively. Generate vectors (or use the render's), then use a roto or mask to reveal blur in specific regions only.
- Check in real time. Some artifacts, such as odd rotational blur or heat-shimmer-like smearing, only show at playback speed.
Managing Motion Blur Across a Whole Script
Motion blur is expensive, so you typically turn it off while animating and back on for preview and render. In Nuke, each Transform has its own controls, with no built-in global switch like the one After Effects offers. In a script with dozens or hundreds of transforms, toggling them one by one is impractical.
Common solutions:
- Use cloned Transform nodes where possible.
- Create a NoOp with custom
mBlurandmShutterknobs, then expression-link every transform'smotionblurandshutterto it. - Write a small Python callback in
menu.pythat links newly created Transform nodes to the controller automatically, so none get missed. - Add a quick project-level checklist item: before final render, search the script for any node with blur at 0.
The failure case to guard against is a single forgotten node that renders sharp while everything around it is blurred. It's easy to miss in a viewer and obvious in a final grade.
Where Styleframe Fits in a Nuke Blur Workflow
AI-generated or stylized footage doesn't arrive with a render-style velocity pass, so the vector-based approach above doesn't apply directly. This is where a controlled generation workflow helps. is an AI video and motion tool built around artist control.
For a Nuke artist, the practical value is in the handoff. Once a generated element is in your script, you treat it the way you'd treat unfamiliar footage:
- Isolate the element with a matte, which also helps the MotionBlur node avoid edge dragging.
- Add motion blur in comp with the MotionBlur node or Smart Vectors, or use Transform-based blur if you're animating the element's position yourself.
- Match the shutter to the plate rather than relying on whatever softness the generation contains.
Keeping blur decisions in Nuke means the look stays adjustable late in the process, and Styleframe stays focused on generating and refining the imagery.
Quick Troubleshooting Checklist
- No blur appears: check that neither
motionblurnorshutteris 0, and that the last transform in the chain isn't overriding it. - Blur is grainy or stepped: raise the sampling rate or Shutter Samples; for painted blur, reduce brush spacing.
- Blur is too long or in the wrong direction: verify vector units, channel assignment, and the VectorBlur multiplier.
- Edges drag or smear: supply a foreground matte or separate vectors for foreground and background.
- Particles look wrong: use render vectors or painted/revealed blur rather than estimating from footage.
- Renders are slow: disable blur while animating, lower samples for previews, and use GPU where supported.
Final Thoughts
Good motion blur in Nuke is mostly about choosing the right source of motion information and matching the shutter. Use transform controls for elements you animate, VectorBlur when you have a clean velocity pass, and MotionBlur or Smart Vectors when you don't. When all of them fall short, painting or selectively revealing blur gives you direct control. Build one global controller early, and you'll avoid most of the headaches that come with larger scripts.