The Pre-Filming Checklist to Run Before Every Swing Video

Most club golfers who film their swing end up with a camera roll rather than a record. Forty clips, shot across eighteen months, from eight different angles, in six different lighting conditions, at three different distances, half of them with a bag strap creeping into frame. Any one of those clips might look fine on its own. Put two of them side by side and you cannot honestly tell whether your hip depth changed or your phone just sat six inches lower that day.

That is the actual problem with filming your swing, and it has nothing to do with your camera. It is a consistency problem. An eight-item checklist, run in about ninety seconds before you hit record, is what converts a pile of clips into a dataset you can compare month to month. Without it you are collecting evidence that proves nothing.

Here is the checklist. Run it in order, every time.

1. Camera height: belt buckle, not eye level

Phone lens at the height of your trouser belt, for both down-the-line and face-on. Not chest, not eye level, not propped against your bag at ankle height.

Why this one goes first: it is the single biggest source of fake swing changes. Raise the camera from belt height to chest height on a down-the-line view and the shaft plane at P2 appears to flatten by a noticeable amount, because you have changed the angle you are looking down the shaft from. Nothing about your swing moved. Sportsbox AI, HackMotion’s video companion, and Swing AI all measure from the pixels they are given, so all of them will dutifully report the change as though it were real.

Use a tripod with a marked leg. A cheap Vanguard or Manfrotto Compact has numbered leg sections; write down the number. If you use a range bucket or a bag as a stand, stop, because you cannot reproduce it.

2. Distance: pace it out, write it down

Down the line: roughly 8 to 10 feet behind the ball. Face on: roughly 8 to 10 feet to the side, level with the ball.

The number matters less than the fact that it is the same number next time. Phone lenses distort near the frame edges, which means that at 5 feet your trail arm looks longer at the top than it does at 10 feet. Pick a distance, pace it (most adult paces run about 30 inches, so three paces gets you close to 7.5 feet), and record it in the same note as the camera height.

A working note looks like this:

Setup profile — 7i, home net
DTL:  tripod leg mark 3, 9ft behind ball, lens on hand path line
FO:   tripod leg mark 3, 9ft to target side, lens level with ball
Frame: head with 1 clubhead of space above, club touching bottom edge at address
Speed: 240fps, 1/1000 shutter (Filmic Pro)

Three lines. Paste it into your phone notes app. It is the difference between a comparison and a guess.

3. Down-the-line alignment: stand on the hand path, not the ball

The classic mistake is putting the camera directly behind the ball on the target line. That gives you a view where your hands and the club overlap your body, and every shaft-plane reading it produces is wrong.

Correct: the camera sits on the line running from the ball through your hands at address, extended backwards. Stand behind, look down the lens, and you should see your hands roughly bisecting your body with the clubhead visible. Sportsbox AI’s own setup guidance pushes you towards this for exactly this reason: their 3D model resolves hand depth far better when the hands are not occluded by the torso for half the backswing.

Test it once, save the alignment, stop thinking about it.

4. Frame rate: 240fps or accept you are guessing at impact

At 30fps, the club travels roughly 4 feet between frames through impact. You will never see impact. You will see a blur before and a blur after, and any tool that claims to measure your shaft lean at impact from 30fps footage is interpolating.

At 240fps (iPhone slo-mo, most recent Android flagships, GoPro Hero 11 and up), you get a frame every 4 milliseconds. That is enough to catch a genuine impact frame on most club golfers’ swings.

Uneekor, Foresight and the better camera-based monitors handle this internally at much higher rates, so if you are filming alongside a GCQuad you are filming for the visual, not the measurement. But if your phone is the only capture device, 240fps is the floor. Not negotiable. And if you run a budget launch monitor like a Garmin R10 or a Rapsodo MLM2PRO, the phone footage is what fills in the gaps in what the radar cannot see, so it needs to be good enough to be read.

5. Shutter speed: freeze the club or the AI cannot find it

Frame rate and shutter speed are not the same thing, and this trips up nearly everyone. Your phone shooting 240fps in automatic mode will still pick a slow shutter in dim light, and you will get 240 blurry frames instead of 30 blurry ones.

Pose-estimation models find your joints reasonably well through motion blur. They find a clubhead badly. Lock shutter to 1/1000 or faster using Filmic Pro (about £15 on iOS), Moment Pro Camera, or the pro mode on most Samsung and Pixel handsets. Outdoors on a bright day you will get away with automatic. Indoors in a net, under a single LED strip, you will not.

Quick check: pause a clip at roughly halfway down. If the clubhead is a smear rather than a shape, your shutter is too slow.

6. Lighting: front-lit, no window behind you

Put the light source behind the camera, not behind you. Backlight turns you into a silhouette, and every pose model working on that footage has to guess where your lead elbow is against a bright background.

This one has a measurable cost. Run the same swing through a joint-detection tool front-lit and backlit and the backlit version routinely produces jittery frame-to-frame joint positions, which shows up as phantom “movement” in depth-based metrics. In a garage or garden net, one £30 LED work light on a stand, aimed at you from beside the camera, fixes it permanently.

Wear something that contrasts with the wall behind you. A navy top against a dark garage wall is a genuinely hard problem for the software; a light polo is not.

7. Background: clear the frame of everything that moves

Trolleys, other golfers, a dog, your bag with a towel flapping. Pose models will occasionally lock onto a second person in frame and swap limbs between them mid-swing. It is a well-known failure and it happens most on busy driving ranges.

Also clear anything club-shaped from the floor behind you. An alignment stick lying in frame gets found as a shaft more often than you would expect.

Thirty seconds of tidying. It costs nothing and it removes the single most common reason a clip comes back with obviously nonsense output.

8. Reference object in frame: give the tool a scale

Put something of known size in the frame, in the same place every time. An alignment stick laid on the target line works well (most are 48 inches). So does a driver headcover, or a strip of tape on the mat.

This does two things. It lets you spot instantly, by eye, whether today’s clip is framed like last month’s. And for tools that estimate real-world distances rather than pixel distances, it gives them a calibration anchor. Sportsbox asks for your height for this reason; a known object in frame gives a second check on whether their scaling is sensible.

Once you have it, the eyeball test takes two seconds: does the stick occupy the same fraction of the frame width as it did in March? If yes, your setup held. If no, something moved.

What the checklist buys you

Take a real case. You film a 7-iron in April and again in September, both at belt height, both 9 feet back, both 240fps at 1/1000, both with the alignment stick in frame. Sportsbox reports pelvis turn at the top going from 38 degrees to 44 degrees, and lead wrist extension at P6 dropping from 23 degrees to 14 degrees.

Those two numbers now mean something. You changed. You can take them to your pro and say: this is what the last five months of work actually did to my sequence, and here is the practice that produced it.

Now take the same two swings with the camera at chest height in April and belt height in September. The September clip shows a steeper-looking shaft, more apparent hip turn, a different apparent spine angle. Every single one of those differences is the tripod. You would spend a month of range time chasing a change that was never there, and that is worse than not filming at all, because it feels like progress.

The related read on all of this, including which tools tolerate which shortcuts, is Filming a Swing an AI Can Actually Read. The short version for anyone who wants it now: HackMotion cares almost entirely about the wrist sensor and treats video as context, Sportsbox needs clean front-lit 3D-friendly footage and will tell you when it is unhappy, and most of the free apps will process anything you give them without complaint, which is exactly the problem.

Running it fast

After a fortnight this takes ninety seconds. The order that works:

That last one, the test swing, catches more problems than the other five combined. Film it, scrub to impact, look at the clubhead. Sharp and framed: proceed. Blurry or cut off: fix it before you hit thirty balls you cannot use.

Build the profile once. Save the note. The next time someone at your club shows you a swing clip and says they think they are getting steeper, ask them what height the camera was at both times, and watch what happens.