Build a Split-Screen Sample With Deliberately Synchronized Actions

Two clips are synchronized in a split-screen sample when a chosen event in one panel lands where you want it relative to a chosen event in the other. That almost never means their first frames line up.

Everything that follows rests on that distinction. You pick the event, place it, then check that each smaller frame still contains the movement that makes the event readable. The layout is the last thing you settle, not the first.

A word about what follows. The example below is a paper construction: frame numbers worked out on a timeline, not frames I watched or built for this article. I have not opened an editing application, rendered a composite, or played anything back. The numbers are internally consistent and you can check them yourself, but they describe a plan, not a result.

Choose the event that relates the two actions

Find the shared anchor. It is the moment a reader can point at: a hand meeting a surface, a glance that arrives, a release, a door's contact, a weight settling. It has to be locatable in both sources, which is a stricter requirement than it sounds.

Take a concrete pairing. In one shot, a hand lowers a ceramic lid onto a pot, and the lid rim meets the pot rim on source frame 48. In the other, a hand presses a stamp onto paper, and the stamp base meets the paper on source frame 30. The shared anchor is first contact. Those two numbers are what you are going to position relative to each other.

Then decide what kind of relationship you want.

• Simultaneous arrival. Both contacts land on the same frame. The viewer reads one unified event happening in two places — the strongest claim, and the one most likely to be misread as evidence that the two actions occurred together.
• Deliberate offset. One contact arrives measurably after the other. The viewer reads sequence, echo, or cause and effect.

Either is a decision. The failure mode is not choosing, and instead starting both clips at zero, which is what a fresh timeline does by default and what most first attempts accept. It aligns the files. It does not align anything a viewer will notice.

A caution before you commit: verify the interpretation of each source. Where is the contact frame exactly — the frame of touch, or the frame where motion stops? Is there a second contact later in the same clip that a careless mark would confuse with the first? What is each clip's frame rate, and does that rate survive the trip into your timeline? A clip shot at 60 frames per second and conformed to a 24-frame timeline gives you a different frame 48 than you marked on set. If the two clips run at different rates, decide which timebase governs before you compare anything, or your offsets will mean different things in each panel.

If you cannot find a shared anchor in both clips, that is generally not a synchronization problem. It is a concept problem. Go back a step and ask what relates these two actions at all, because no amount of timeline work will supply a relationship the footage does not have.

Place the sources on a common timeline

One composition, one timebase, both sources in it. Adobe's After Effects help page on layer properties — the one recorded source for this piece, checked 9 September 2026 — describes adjusting a layer's position, scale, opacity and temporal properties, and notes that scaling and rotation use the layer's anchor point. That is enough to establish the construction route: you move a layer in time and in frame through the same properties panel.

What that source does not do is tell you anything about whether a particular pairing is meaningful or correctly timed. It records operations, not relationships. Treat the control-level detail as something to confirm in your own installed version; I have not walked through the current interface for this article.

Now the arithmetic. A 24-frame timeline, with the anchor placed at frame 36 — one and a half seconds in. Clip A is five seconds, 120 frames, contact at source frame 48. Clip B is four seconds, 96 frames, contact at source frame 30.

Getting clip A's contact to frame 36 means its first twelve frames fall before the sample begins, so you trim them. Clip B needs to start six frames in.

That produces two side effects you should notice rather than discover later. The panels do not enter together: A is present from frame 0, B arrives at frame 6. And they do not leave together either — B's last frame sits at 101, while A runs to 107.

Keep the original clips accessible while you work. Either duplicate them below the visible layers and disable the copies, or leave the source media untouched in the project and treat the composite as a separate construct. The check you will need to perform repeatedly is whether the action in the composite still matches the action in the original, and you cannot perform it against a trimmed layer whose head you have already removed.

Frame each action for its smaller area

A divided frame halves the space each action gets, so the crop is where the anchor either survives or quietly disappears.

Suppose the composite is 1920 by 1080 and each panel is half of it: 960 wide by 1080 tall. A source clip is also 1920 by 1080. At 100 percent scale, the panel shows a 960-pixel-wide window onto a 1920-pixel-wide image, and the only question is where that window sits. Centred, it shows source columns 480 to 1440.

Check where the meaningful pixels fall. The lid meets the pot at roughly column 1180, comfortably inside the centred window. But just before contact, the hand holding the lid is at roughly column 1620, descending. Outside the window. So the panel shows a lid landing on a pot with no hand attached to it: the preparation is cropped away, and the action loses its cause.

The fix is a horizontal move of the layer, not a rescale. Shift the layer 220 pixels left — position x from 480 to 260 — and the window covers source columns 700 to 1660, still 960 wide. Left is the direction that reveals the columns to the right of the centred window, which is where the hand is; shift right instead and the window covers 260 to 1220, with the hand still outside. The hand at 1620 is inside, and the contact at 1180 remains inside. One number changed, and the gesture became legible end to end.

Two things to keep in mind when you do this.

The first is that scale and position are entangled through the anchor point, which is exactly what the recorded source notes. If you rescale a layer, the framing moves around the anchor rather than around the panel centre, so a scale change can slide your carefully positioned contact out of frame. Move position first; only rescale if position alone cannot do the job.

The second is that you should inspect the whole gesture, not the contact frame. A crop that frames the contact beautifully can still amputate the approach. Scrub backwards from the anchor for at least as long as the preparation lasts — here, about fourteen frames — and forward past the contact far enough to see whether the hand withdraws out of frame or lingers. The withdrawal often reads as part of the same action, and cutting it off mid-motion leaves the panel looking truncated.

There is an alternative worth naming: instead of cropping, scale each source down to fit inside its panel. That keeps the whole action visible and adds letterbox space around it. It is a legitimate choice, usually a more neutral and less cinematic one, and it costs you the tight framing. Decide; do not arrive there by accident.

Make entry, exit, and sound deliberate

The simultaneous interval is the part everyone thinks about. The entry and exit are where the sample most often goes wrong.

Entry. A enters at frame 0 and B arrives at frame 6 — a quarter of a second later. If B simply appears, that is a visible pop in the right half of the frame, and it draws attention to the mechanics rather than the relationship. Three reasonable answers: trim A's head by six more frames so both enter together and move the anchor accordingly; bring B in under a dissolve or a wipe so its arrival reads as intentional; or extend the sample backwards and let one panel be alone for a beat before the other joins it, which stages the pairing as an event. Any of these is defensible. Leaving a quarter-second pop in place because you did not notice it is not.

Exit. B finishes at frame 101 and A runs on to 107. Six frames, a quarter of a second, of a half-empty frame. Hold B's final frame so both panels resolve together; let B's area fall to black; cut the whole composite at 101 and lose A's last six frames; or keep A alone for those six frames as a deliberate coda. Again, choose. The one option that is not a choice is letting the sample end wherever the longer clip happens to stop.

Sound. In the aligned version both contacts land on frame 36. If both source tracks stay audible, you get two transients on one frame. Depending on the sounds, that either fuses into a single heavier impact — which strongly implies one cause producing both images — or it turns to mush.

Your options are honest and limited. Keep both, and accept that you have written a unison. Drop one, and let a single sound carry both images, which is the cleanest way to say "these two things are one thing." Keep both and offset the sounds even when the pictures are aligned, which is a legitimate trick and one you should disclose if it matters. Or go silent and let the images argue alone.

Note what each choice does to the second version, where B is moved later. At eight frames of offset — one third of a second at 24 frames — the two impacts stop fusing and become separately audible. That is not a small change. The same two recordings, moved eight frames apart, stop saying one event and start saying two events, in order. If your treatment depends on the first reading, the offset version is not a variant of it. It is a different argument.

Which brings up the disclosure point. Synchronized images made in an edit are a constructed relationship. A sample that presents two separately recorded actions landing on the same frame is asserting a correspondence you built, not documenting one that occurred. If the sample is going to be read as evidence that anything happened simultaneously in the real world, say in the treatment notes that the alignment is deliberate. This costs you nothing and prevents the most damaging misreading of the work.

Compare aligned and offset versions in playback

Build the offset version by moving one source. Nothing else changes: same crops, same scale, same two clips, same anchor choice. Move B eight frames later. Its entry shifts from frame 6 to 14, its contact from 36 to 44, and its final frame from 101 to 109.

That last number is the consequence people miss. In the aligned version, A outlasts B by six frames. In the offset version, B outlasts A by two, and the composite has to run two frames longer to contain it. Changing the temporal relationship between two events changed which panel finishes first — which is exactly why the exit decision has to be made again, not inherited from the previous version. Compare complete alternatives under the same conditions, or you are comparing two different samples.

Then watch. Play the whole passage through once before you inspect anything, because the anchor frame in isolation tells you what the relationship is but not how it feels. A freeze on frame 36 or 44 shows two panels at a moment; it cannot show you whether the second contact reads as an echo, a reply, or a mistake. Only playback can.

After the full pass, go to the anchor and step frame by frame. In the aligned version, both actions arrive together and the eye has to choose where to look — which is a real problem with simultaneous contact in two panels, and worth knowing before you show the sample to anyone. In the offset version, the lid lands, then a third of a second later the stamp presses, and the viewer's attention gets handed from one panel to the other. That handing-over is what the offset buys you.

What it costs is the claim of simultaneity. Say that plainly in your notes rather than deciding that tighter synchronization is inherently better. It is not. It is a different proposition about what the two actions have to do with each other.

One deliverable consideration. A treatment often wants a still, and a frame grab at the anchor is a good one — divided frame, both actions at their peak, legible in a document. Keep it, and be clear about what it is. A still can show that both contacts exist and that both gestures fit inside their panels. It cannot show order, lag, entry, exit, or sound. Anyone reading a still will assume the two events happened at the same instant unless told otherwise.

What the sample has to contain when you are done

A chosen anchor, named and located in both sources. One composite timeline where that anchor sits at a stated frame, with each panel's entry and exit decided rather than inherited. A crop for each side that preserves the movement leading into the anchor, not only the anchor itself. An explicit audio choice, documented. And two versions — aligned and offset — that explain, by their difference, why you chose the one you did.

The layout is the least interesting part of the whole operation. A symmetrical divide looks tidy in a document and proves nothing about time. The sample earns its place in a treatment when it demonstrates that you know which two moments are supposed to relate, and how far apart you want them.

Frequently asked questions

What counts as synchronization in a split-screen sample?

A chosen event in one panel lands where you want it relative to a chosen event in the other. It does not mean the clips' first frames line up. The shared anchor might be first contact, a glance, a release, or a door's contact, and you decide whether the relationship is simultaneous arrival or a deliberate offset.

What should I verify before trusting an anchor?

Check where the contact frame actually is: the frame of touch or the frame where motion stops. Look for a later second contact that could be confused with the first. Check each clip's frame rate and whether it survives conforming; if rates differ, decide which timebase governs before comparing offsets.

How do I crop each panel without losing the action?

Inspect the whole gesture, not just the contact frame. A centred crop can keep the contact but cut off the hand that caused it. Shift position first; rescaling interacts with the anchor point and can move the contact out of frame. Scrub backwards through the preparation and forward past contact to see whether the withdrawal matters.

How does an eight-frame offset change the sample?

Moving one source eight frames later shifts its entry, contact, and final frame; in the example, the other clip then outlasts it by two frames and the composite runs longer. The two impacts also stop fusing into one event and become separately audible, which reads as two events in order. The exit decision has to be made again.

What must the finished sample contain?

A named anchor located in both sources, one composite timeline with the anchor at a stated frame, deliberate entry and exit choices, crops that preserve movement into the anchor, and an explicit audio choice documented. Build aligned and offset versions so their difference explains the choice. If the sample could be read as evidence that two actions happened together in reality, disclose that the alignment is constructed.

Further reading

• How to Make a Reference Contact Sheet From Video Stills — If a reviewer will get a frame grab of the anchor, this explains how to record the still's clock and position so the moment stays findable.