How to film a magic trick video with both iPhone cameras
A magic trick posted online meets one reflexive objection before anything else: that it was cut together, that the reaction was filmed separately from the trick, that the two halves never actually happened at the same time. Recording the front and back camera at once answers exactly that objection, because the format has no separate halves to begin with — the trick on the back camera and the spectator’s face on the front camera are drawn into one frame together, in real time, and that composited frame is what gets written to the file. There is no edit point between them to be suspicious of, because there was never a cut to make.
That is a genuinely different starting point from most of what this site covers. A vlog or a tutorial benefits from a single take; a trick reveal depends on it, because the credibility of the whole clip rests on the audience being able to tell that nothing was spliced in after the fact.
What an unbroken take actually proves — and what it does not
Be precise about what this buys you. A composited recording proves that the two images you are looking at came from the same real second, captured by two cameras running at once on one device — it rules out the specific, common accusation that the reaction was filmed on a different day and stitched to the trick afterward. It does not prove the trick itself was not staged, does not prove the spectator was not in on it, and does not stop someone from re-encoding or re-cutting the finished file later on a different device. It removes one objection cleanly; it is not a certificate of authenticity for everything else a skeptic might raise.
That narrower claim is still worth a lot. "That was obviously two clips" is the single most common dismissal a trick video gets, and it is the one this format is structurally unable to deserve — there was only ever one clip, composited as it was captured, with the same continuous audio track running under both halves.
The trick fills the frame; you are the small window
Picture-in-picture is the right default here, for the same reason it is the default for a reaction shot generally: the thing that needs to read clearly — hands, cards, a coin, whatever the trick is actually built on — gets the full frame, and your face sits in the inset, small but present for the reaction it is there to capture. A trick performed with a partner in the same room, where both of you need to be seen at full size, is the exception that calls for split screen instead — but a left-right split trims roughly a quarter of the source frame from each side on a vertical canvas, so keep the sleight of hand centred rather than out toward the edge of your half.
This is worth deciding before anything else about the shot, because it changes where you physically stand and how wide the gesture can be. A trick that is fine to perform close to the lens in picture-in-picture may need to be performed further back, or smaller, to survive being cropped into half a split screen.
Lock the framing and the lens before you say "watch this"
Which lens the back camera is using, and the resolution, are both fixed the moment recording starts — there is no crossing to a different physical lens mid-take, only pinch-zooming within whichever one you picked. If the trick needs the camera close enough to read a card face or a coin, or far enough back to show that your hands never leave the frame, that distance has to be right in the first take, because there will not be a natural pause to correct it once the trick is underway.
Rehearse the exact framing once with the phone rolling on nothing in particular — no trick, no spectator — just to confirm the crop, the distance and the lens are all doing what you expect. Layout, the picture-in-picture corner and its size can still be adjusted while a take is running if something shifts, but the lens and the resolution cannot, so get those two settled before anyone is watching.
One microphone, pointed at the reaction that pays the trick off
A dual-camera recording carries a single audio track from the phone’s own microphone, not one track per camera, and it picks up whatever is physically closest to it — which, held by the performer, is usually the performer’s own patter rather than the spectator’s reaction. The line that actually sells a trick is rarely anything you say; it is the spectator’s gasp, the "wait, what," the request to see it again. If that line matters more than your own narration, angle the phone or position yourself so the spectator ends up closer to the microphone than you are, rather than trusting a single track to pick up the right voice on its own.
A spoiled take cannot just be reshot on the same person
There is no pause button in the middle of a dual-camera recording — stopping and starting again is the only way to break up a shoot, the same as it is for every other subject on this site. For most subjects that is a minor inconvenience: press stop, reset, go again. A magic trick is the exception, because the moment the same spectator sees the method go wrong, or even hesitates in a way that reveals a tell, that specific performance is burned for that specific person — recording it again does not get back the genuine surprise the first attempt was supposed to capture.
The practical fix is to treat rehearsal and recording as separate activities entirely, rather than using the camera as the place you find out whether the trick works. Get the trick itself reliable off camera first, decide the framing and lens as above, and only press record once you are confident the first take will also be the only take you need — with a fresh spectator, and ideally a fresh angle on the trick, held in reserve for the rare case where it has to be attempted twice.
Common follow-up questions
Does recording this way prove the trick was not faked?
It proves the trick and the reaction happened in the same real second on the same device, since both are composited into one frame as they are captured — there is no separate clip to have spliced in later. It does not prove the trick’s method, and it does not stop the finished file from being staged or re-cut somewhere else afterward. It answers the "that was obviously two clips" objection specifically, not every possible objection.
Should I use picture-in-picture or split screen for a trick?
Picture-in-picture, in most cases. The trick needs to read clearly at full frame, and your face only needs to be visible, not equally sized — that is exactly what the inset is for. Reach for split screen only when a second performer genuinely needs equal billing, and keep the sleight of hand centred, since a left-right split crops roughly a quarter of the frame from each side.
What if the trick goes wrong while I am recording?
Stop and start a new take — there is no way to pause and resume the same file. Unlike most subjects on this site, though, a botched attempt usually cannot be repeated in front of the same spectator without giving something away, so treat the recording as the performance itself: rehearse off camera until the trick is reliable, and only press record once, rather than using the camera to find out whether it works.
Want to just do this?
DualCam records the iPhone front and back cameras at the same time and writes one finished MP4 while you shoot. Free, no account, no ads, nothing leaves the phone.