How to make a split-screen video of two people talking, from one camera

A podcast is usually one camera on two people. Cut it down to a vertical clip and you lose one of them: the frame is too narrow for both, so the edit shows whoever is talking and the other person becomes a voice. A split screen keeps both faces, one above the other, from that same single shot. The whole trick is cropping each half so hard that only one person survives in it.

One camera, two crops

Nothing in a split screen separates the two people. Both panes show the same wide shot at the same instant; each is zoomed and panned until it holds one face. The failure everyone hits first is not zooming enough. Put a 16:9 recording into a pane that is 9:16 and full height and the pane already sees roughly 63 percent of the frame’s width, which is enough for both people to fit no matter where you pan it. Zoom first, by around 1.4 times for a normal two-shot, and the visible slice drops to about 45 percent; only then does panning one pane to the left end and the other to the right land on one face each. People sitting far apart need less zoom; people close together need more, up to 1.6 or 1.8 times.

The rule, if you are doing it by eye: zoom until one person fits, then pan to the end. Keep each face centred in its half so the seam between the panes never cuts a chin.

When the split earns its place

A split screen that runs for the whole clip is the boring version of it. It flattens everything to the same size and gives the viewer nothing to look at when only one person is talking. Split on the exchange, not on speech: the moments when both faces are worth seeing at once.

• A question and its answer landing within a couple of seconds of each other.
• A reveal and the listener’s reaction.
• Two people talking over each other, or sharing the same confusion.
• The punchline and the take.

A monologue stays solo; the second pane would have nothing to show but someone nodding. Windows of three to five seconds, starting a fraction before the first word of the exchange, cut in and out four to six times across a clip, read as punctuation rather than a layout. If the clip ends on a reaction, let the last window run to the end so the frame does not snap back to a single face on the final beat.

And put the person the clip is about on top. A phone shows the top of the frame first as it scrolls into view, so the top pane is the thumbnail.

How to ask

On a podcast or an interview, Narrative asks three things before it cuts: the aspect ratio, the speaker layout (split screen, only whoever is speaking, or a mix of the two), and whether B-roll goes over the talking. The mix is almost always the right answer, and you can say exactly which mix:

• “Split screen on the back-and-forth, solo framing the rest of the time.”
• “Both of them on screen when she asks the question and he answers; him alone for the story.”
• “Keep the host on top throughout.”
• “Split only for the last ten seconds, on the reaction.”

Cut the part about the first job into a 45-second 9:16 clip. Split screen when they talk over each other and on the two questions; him alone otherwise. Captions on.

The editor finds the exchanges from the speaker turns of the transcript: an exchange is a run of short turns alternating between speakers, and long single-speaker turns are the monologues left solo. It measures where the faces are in the shot to compute each pane’s crop and zoom instead of guessing, and if the camera cuts during the clip, it changes the crops at the cut rather than letting a face drift out of its pane. Then it reads the crops back at a moment inside a window, to confirm each pane shows the slice it meant, before it replies.

What you get on the timeline

Each split window is its own item on a Split lane, anchored to the clip it sits in. Trim one shorter, slide it a second later, delete the one that did not earn its place; the others stay where they are, and between windows the clip plays in its ordinary single-face framing. Because the panes are clones of the clip rather than copies of the footage, the clip’s own sound is untouched: no doubled voices, nothing to re-sync when you move a window.

Captions in the podcast style (large, bold, the current word popping) go on as usual; the captions post covers where they sit and why. For the reframing underneath, when it is a single face rather than two, the 9:16 reframing post is the one to read.

Mistakes to look for

• Both panes show both people. The zoom is too low. Nothing you do with the pan will fix it.
• The split is on for the whole clip. Cut it back to the exchanges and the clip gets its rhythm back.
• A face on the seam. The crop is panned too far; bring the face toward the middle of its pane.
• The guest on top when the clip is about the host. Swap them. The top pane is what the feed shows first.
• A split that starts after the first word. Open it a fraction early, so the second face is already there when the line lands.

Try it

Upload the wide shot of a two-person recording, ask for a clip, and answer “split screen on the back-and-forth” when it asks about the layout. Creating an account takes a minute; the first edit starts a three-day free trial. The web editor runs on a computer; on a phone, the iPhone app is the way in. If you have a longer episode and want several clips out of it, start with the podcast-to-clips post.