Motion Blur for VFX

Motion blur is the streaking that appears when something moves while the shutter is open. On a real camera it is an optical fact of exposure time. In VFX it is something you must match, not invent: wrong blur is one of the fastest ways to make CG, plates, and composites look fake.

This lesson covers how film and pixels accumulate light into blur, the 180° shutter convention (including 24 / 23.976), rolling shutter, when to break the rules, and how to actually produce blur in 2026 — primarily real 3D motion blur in the renderer, with motion vectors as a limited fallback.

How exposure creates blur (film and pixels)

A camera frame is not an instant snapshot. It is a record of light collected over a stretch of time.

Integration, not a single moment

While the shutter is open (or while a sensor is collecting charge):

1. Light from the scene hits the film grain or the photosite (pixel).
2. That light is added up — film density builds; a digital pixel’s well fills with charge proportional to photons.
3. If anything in the scene (or the camera) moves during that time, light from different positions of the same object lands on different grains / pixels — or smears across several.
4. When the shutter closes (or readout finishes the exposure), you get one image: the average (integral) of everything that was visible over the open interval.

So motion blur is not a filter someone invented for “cinematic feel.” It is the natural result of:

pixel value ≈ ∫ light(t) dt (from shutter open → close)

A still object paints the same place the whole time → sharp. A fast object paints a trail → streaked. A pan moves the whole world across the frame → background blur.

Same physics, two media

For VFX matching you care about how long integration lasted and when each part of the frame was integrating — not whether the medium was silver halide or CMOS.

How much blur you get

Blur length on the image is roughly:

blur length ≈ image-plane speed × open time

• Open time comes from shutter angle + frame rate (or from a digital shutter speed like 1/48 s).
• Image-plane speed is how fast the subject (or camera) moves across the frame — not just world-space mph. A distant jet can look almost sharp; a near hand can smear hard at the same shutter.

That is why matching “180°” is not enough if the CG subject’s speed and distance don’t match the plate. Same open time, different on-screen velocity → different streak length.

What a “sharp” frame really is

Even a “frozen” action still integrated light for some non-zero time. A tighter shutter (90°, 1/1000 s, etc.) just makes that window short enough that the trail is only a few pixels or less — so the eye reads it as sharp. Nothing in production is a true Dirac delta at one instant unless you are in extreme high-speed / strobe territory.

Global shutter vs “the whole frame at once”

Idealized film / global-shutter digital: every part of the frame integrates over roughly the same time window. The whole image is one coherent exposure. Motion blur is then mostly streaking along motion paths, and the geometry of the world stays consistent within that frame (aside from the blur itself).

Many cinema cameras and almost all film behave this way (or close enough for matching). Rolling shutter (next section) breaks the “same time for every pixel” assumption.

The 180° shutter, fully explained

What a shutter angle is

Film cameras (and the digital cameras that imitate them) often describe exposure time with a shutter angle rather than a shutter speed like “1/48 s.”

Imagine a spinning disc in front of the film gate:

• A full circle is 360°.
• Part of that disc is open (light passes) and part is closed (light blocked).
• The size of the open sector is the shutter angle.

While the disc spins once per frame, light only hits the film during the open sector. So:

shutter open time = (shutter angle / 360°) × (1 / frame rate)

The film default: 180° ≈ shutter 0.5 at 24 / 23.976

For most narrative film and cinematic VFX, you can assume this stack unless the camera report says otherwise:

Those labels are different units for the same convention:

180° ⟺ 180/360 = 0.5 ⟺ open for half of each frame

At true 24 fps:

At 23.976 fps (often written 23.98):

Important for VFX: the Nuke / renderer control is still 0.5 at either rate. Shutter fraction is “how much of the frame,” not a magic number tied only to 24.000. The absolute open time shifts by a tiny amount with fps; the look of a 180° shutter does not.

In practice:

• Comp and CG default to 180° / 0.5 for film plates.
• You set the project / plate fps to 24 or 23.976 to match the camera and editorial.
• You do not change shutter to 0.5 only “because it’s 24” and to something else “because it’s 23.976.” Same 180° rule; different timeline rate.

That is the industry default “cinematic” motion blur.

Why 23.976 instead of 24?

24 fps is the classic sound-era film rate: enough frames for smooth motion and lip sync, economical on film stock, and locked in as “the movie look.”

23.976 fps is not a different creative choice. It is 24 fps adjusted for NTSC video math.

Short version of the history:

1. US/Japan broadcast video (NTSC) ran at 29.97 frames per second, not a clean 30 — a consequence of adding color to the old black‑and‑white 60 Hz system while staying compatible with existing TVs.
2. To put 24 fps film onto that 29.97 video world, facilities used 3:2 pulldown (telecine): film frames are repeated in a 3-field / 2-field pattern so 24 film frames map into 30 (really 29.97) video frames.
3. For that relationship to stay exact over long runtimes, film was treated as running at:

24 × (1000/1001) = 24000/1001 ≈ 23.976 fps

rather than a pure 24.000. That is the same 1000/1001 slowdown that turns 30 → 29.97.

So:

For motion blur: 23.976 vs 24 barely changes open time at 180° (about 20.85 ms vs 20.83 ms). What does matter is matching the plate’s actual fps in your comp script, retime, and 3D scene so frame timing, audio sync, and editorial lengths stay correct. Shutter stays 180° / 0.5 in the normal case.

If a camera report says “23.976, 180°,” set fps to 23.976 and shutter to 0.5 — not “something in between” because the number looks weird.

Why Nuke defaults to 0.5

In Nuke (and many other tools), shutter is often expressed as a fraction of the frame:

• shutter = 0.5 → open for half the frame → 180° shutter
• shutter = 1.0 → open for the whole frame → 360° shutter
• shutter = 0.25 → open for a quarter of the frame → 90° shutter

Conversion:

shutter angle = shutter fraction × 360°

shutter fraction = shutter angle / 360°

99% of the time, leave Nuke’s default of 0.5 alone on film work at 24 or 23.976. That is the 180° convention. Changing it without a plate or camera report that says otherwise will usually fight the footage.

Other common angles (same math)

At other frame rates the angle still means “fraction of the frame interval”:

• 180° @ 48 fps → open 1/96 s
• 180° @ 60 fps → open 1/120 s

The amount of blur on screen also depends on how far the subject moves during that open time, not only on the angle.

Why 180° became the rule

Historically, a half-open rotating shutter was a practical mechanical default. Visually it stuck because it produces motion blur that feels natural for narrative film: moving objects soften enough to read as continuous motion, without turning every pan into a smear.

People often say it “matches human vision.” That is only a rough analogy. The useful VFX takeaway is simpler:

Most narrative plates are shot near 180°. Match that unless the plate or brief says otherwise.

Exposure is usually not controlled by changing the shutter

If a cinematographer needs a darker or brighter image, they typically do not throw away the 180° look. They use:

• ND filters (cut light without changing shutter angle)
• Lighting
• ISO / EI
• Aperture (T-stop / f-stop)

Changing shutter angle does change exposure (longer open time = more light), but it also changes the motion character of the shot. That is a creative choice, not the default exposure tool.

Shutter phase (centering the blur)

Shutter angle answers “how long?” Shutter phase answers “when, relative to the frame?”

Common conventions:

• Centered shutter — exposure straddles the frame time (half before, half after). Often what CG and Nuke assume.
• Start / end shutter — exposure begins or ends on the frame sample.

If CG blur feels slightly “offset” from the plate even at 180°, check phase and whether the renderer’s shutter interval matches the camera report — not only the angle.

Rolling shutter

Shutter angle / open time answers: how long did light accumulate? Rolling vs global shutter answers: did the whole frame accumulate over the same interval, or row by row?

Global shutter (the simple mental model)

With a global shutter (film with a rotary shutter, some digital cinema cameras, many CG renders):

• All pixels (or the whole film gate) start and stop integrating together.
• One frame = one time window for the entire image.
• Moving objects streak, but vertical poles stay straight, wheels don’t “lean,” and a fast pan doesn’t skew architecture into a parallelogram.

That is the model most path-traced motion blur assumes unless you deliberately simulate rolling readout.

What rolling shutter is

Most CMOS sensors (phones, many mirrorless / cinema cameras, a lot of production glass) do not expose every row at the same instant. They read the sensor from top to bottom (or bottom to top): each row starts and/or finishes its integration slightly later than the row above.

Think of a slit sweeping down the frame:

time →

row 0  [====expose====]
row 1    [====expose====]
row 2      [====expose====]
row 3        [====expose====]
...

Each row still accumulates light over its own open time (so you still get normal motion blur / exposure). But the center of that exposure window is offset in time from top to bottom. The full “rolling readout” duration can be several milliseconds — sometimes comparable to the open time itself.

How that looks on set (and in plates)

When subject or camera motion is fast relative to the readout:

Rolling shutter is not the same as motion blur. You can have:

• Long open time + global shutter → smooth cinematic streaks, no skew
• Short open time + strong rolling readout → sharp but skewed frames
• Both → smeared and sheared (common on phones and some cameras)

Still light accumulation — just not simultaneous

Per photosite, the story is unchanged: charge builds while that site is integrating. Rolling shutter only changes the schedule:

1. Row i opens (or starts its effective window).
2. Photons add into those pixels for ~open time.
3. Row i is read / closed.
4. A short time later, row i+1 does the same for a later slice of the world.

So blur is still ∫ light(t) dt per pixel — but t is not shared across the frame. Matching only “180° global” CG to a heavily rolling plate will get streak length roughly right and geometry over time wrong.

VFX implications

1. Know the camera. ARRI, RED, Sony, etc. differ in readout behavior; phones are often worst. Check the camera report / model, not only shutter angle.
2. Tracking & layout — rolling plates can need rolling-shutter-aware solvers or compensated tracks. Straight 3D trackers assume global time.
3. CG integration — some pipelines render global motion blur and treat mild rolling as acceptable; high-end or phone/action work may need rolling shutter in the render, a warp to match the plate, or both.
4. Comp “fix” — CornerPin / grid warps, RollingShutter nodes, or optical-flow un-roll tools can reduce skew for 2D work; they don’t replace correct 3D exposure for hero CG.
5. Don’t confuse with 180°. Leaving Nuke at 0.5 matches open fraction. It does not automatically give you rolling readout. Those are separate knobs in the real world (and in careful CG).

Quick contrast

Exceptions to the 180° rule

1. Plates intended for heavy retiming

If a shot will be slowed down a lot in post, production may shoot a tighter shutter (e.g. 90° or less) so each frame is sharper.

Why: it is relatively easy to add synthetic motion blur later (optical flow, vector blur, retime tools). It is nearly impossible to remove real motion blur cleanly. Blur already baked into the pixels destroys detail you would need for a clean retime.

Tradeoff: a tighter shutter can look jittery or hyper-real if left at native speed — which is sometimes desirable for action (see below).

2. Stylized / intense action

A shorter shutter (smaller angle) freezes motion more. Action can feel snappier, more violent, or more documentary/combat-like. Saving Private Ryan–style shutter choices are a famous extreme of this idea: less blur, more raw impact.

3. Dreamy or smear-heavy looks

Larger angles (or long digital shutters) increase blur. Use sparingly and deliberately; heavy blur can also hide detail you need for tracking and keying.

4. High frame rate capture

HFR (48, 60, 120 fps, etc.) changes how much movement occurs per frame. Matching blur still means matching angle (or open time) and subject speed, not blindly copying “Nuke 0.5” without understanding the plate.

How to actually do it (2026)

Default: proper 3D motion blur in the renderer

In 2026, for almost all feature and serious TV CG, turn on real 3D motion blur in the render engine and ship beauty with blur already correct. Do not plan on “fixing it in comp” as the primary path.

That means Arnold, RenderMan, Cycles, V-Ray, Karma, Maya Software-era successors, Clarisse, Hydra delegates — whatever your show uses. They all share the same idea:

1. Set scene / camera fps to the plate (24 or 23.976, etc.).
2. Set shutter to 180° or fraction 0.5 (and match phase if the tool exposes it).
3. Enable motion blur on camera, transforms, and deformation as needed (skin, cloth, sim).
4. Let the renderer sample the scene across the open shutter interval and integrate lighting the way a real exposure would.

What you get that 2D cannot honestly do:

• Curved blur paths (not only straight smears)
• Motion through occlusion, self-intersection, contact
• Transparency, volumes, hair, particles under a true multi-sample exposure
• Speculars and shadows that move during the shutter
• Correct treatment of rotation and of motion toward / away from camera

That is the correct workflow. Comp’s job is usually to integrate that already-blurred CG with the plate — grade, holdouts, interactive light — not to invent hero motion blur from a sharp render.

Exact UI names differ. The checklist does not: fps match → 180° / 0.5 → enable 3D moblur → render.

Practical setup checklist (3D)

1. Camera report — fps, shutter angle, phase notes.
2. Project fps — 24 or 23.976 (or whatever the plate is), not a random default.
3. Shutter — 180° / 0.5 unless told otherwise.
4. Motion samples — animation (and cache) must exist across the shutter, not only on whole frames.
5. Lead-in / lead-out — keys past the first and last frame of the shot so end frames still blur correctly.
6. Render beauty with moblur on for hero assets. Turning it off “to save time” and hoping comp will save you is how you get soft failures in finals.
7. Holdouts / deep / IDs — still work with 3D moblur; plan AOVs knowing edges are already soft.

Animation pro-tip

When setting animation keys, extend motion past the first and last frames of the shot (or of the render range).

If an object is still on frame 1001 and only starts moving on frame 1002, the first moving frames will not have continuous, correct shutter samples. Lead-in and lead-out motion keeps blur continuous at the cut.

Also ensure camera and object motion samples exist across the full shutter window (not only on integer frames).

The other option: motion vectors (screen-space only)

You can render sharp (or lightly blurred) beauty plus a motion vector pass, then blur in 2D — Nuke VectorBlur, After Effects, Blender vector blur, etc.

That is still valid for:

• Cheap TV / student / previz budgets
• Small elements and support CG
• Salvage when a vendor delivered sharp plates and there is no re-render
• Touch-ups and very mild motion

It is not a full substitute for 3D shutter sampling on hero film work.

What a motion vector actually is

A motion vector AOV stores, per pixel, roughly where that surface is moving in screen space — e.g. offset in X/Y (pixels or NDC) between times bracketing the frame.

Comp then smears the color along that 2D vector. That is a post-process approximation of exposure, not a re-integration of the 3D scene.

Foundry’s VectorBlur overview: https://learn.foundry.com/nuke/content/comp_environment/3d_compositing/adding_motion_blur_vectorblur.html

Hard limits — keep these in mind

Vector blur is screen-space only. It does not re-trace the scene. In particular it struggles or fails when:

Rule of thumb: if the interesting motion is lateral in frame and fairly rigid, vectors can look acceptable. If the motion is rotation-heavy or in Z (to/from camera), use real 3D motion blur.

3D vs vectors (decision table)

Comp habits that still help

• Even with 3D moblur, you may use light vector or optical-flow blur for 2D cards, grain, or retime.
• Never assume you can deblur a plate to a sharp frame and rebuild — you generally cannot.
• If you must use VectorBlur, match shutter to 0.5 / 180° and check edges against the plate; don’t crank “blur amount” until it “feels right” without measuring.

Practical matching checklist

When integrating CG or a retime into a plate:

1. Read the camera report — fps (24 vs 23.976), shutter angle (or shutter speed), camera model, and any notes about creative shutter or rolling.
2. Default to 180° / 0.5 if undocumented narrative footage — at 24 or 23.976; match fps separately from shutter.
3. Match open time, not a random “blur amount” slider. Convert angle ↔ fraction ↔ seconds if tools disagree.
4. Match shutter phase if blur feels time-offset from the plate.
5. Check for rolling shutter on fast motion, handheld, props, flashes — skew/jello is not fixed by changing Nuke’s 0.5.
6. Do it in 3D — Arnold / Cycles / RenderMan / etc. with shutter 0.5. Vectors only for elements, temp, or salvage.
7. If using vectors, remember: screen space only — bad at rotation and to/from camera motion.
8. Don’t strip real plate blur expecting to replace it cleanly — plan acquisition for retimes instead.
9. Keep animation moving through the shutter on first/last frames.

Quick reference

Summary

• A frame is integrated light: film or pixels accumulate photons over the open window. Motion during that window is what we call motion blur.
• 180° shutter means open for half of each frame. In Nuke that is shutter = 0.5. That is the default for most film work.
• At 24 fps, open time is 1/48 s. At 23.976, it is almost the same (~1/47.95 s). Shutter fraction stays 0.5; only the project fps changes.
• 23.976 exists because of NTSC / 29.97 video math (24 × 1000/1001), not because cinematographers wanted different motion blur.
• Rolling shutter staggers when each row integrates. You still get accumulation blur, plus skew / jello / bent motion. That is separate from setting shutter to 0.5.
• Leave shutter at 0.5 unless the plate, camera report, or creative brief says otherwise. Match 24 vs 23.976 and rolling vs global from the report / camera, not from guesswork.
• Cinematographers usually protect the 180° look and control exposure with ND, lighting, ISO, and aperture.
• Break the 180° rule for retimes (tighter shutter, sharper frames) or for a deliberate jittery/smeared style.
• How to do it in 2026: enable proper 3D motion blur in Arnold, Cycles, RenderMan, or whatever the show uses — fps match, shutter 0.5, render it for real.
• Motion vectors are still possible in comp, but they are screen-space only: they do not handle rotations or movement toward/away from camera well. Use them for elements, temps, or salvage — not as the default for hero film CG.
• Keep animation continuous through the shutter on the first and last frames of a range.