Film yourself squatting from the side. You get depth against parallel, torso angle, tempo, and whether the last reps still look like the first. A back-view clip answers a different question — symmetry — and we do not grade it, because 2D video is not a diagnosis for hip drop or knee cave.
Side-on tells you how deep: depth against parallel, torso angle in the hole, whether you stand fully between reps, and how consistent the set is. That is the clip we score. A camera from behind would speak to whether one hip drops, you shift to one side, or the knees track in. Those are a different plane. A side-on phone cannot answer them, so they stay off the card rather than being guessed from the wrong angle. The back clip is a screen, not a score — and the reasons sit at the bottom of this page.
How to film the side clip
Use the 1x lens, never 0.5x. Ultrawide bends straight lines worst at the frame edges, which is exactly where the joints we measure sit. Stand 10 to 12 feet back — distance kills perspective error faster than anything else you can do. Whole body in frame, feet included. Five or more reps. The height of the camera is the most common mistake. Put the lens level with your hip and knee at the bottom of the rep, about two feet off the floor: a chair seat or a stack of books. At the bottom your hips travel back and your knees travel forward, so they sit at different distances from the lens. Filmed from chest height that difference is projected into an apparent depth that is not real, and it makes you look shallower than you are. At knee height the error nearly vanishes. Square to your side, not at an angle. Put a count on it: three seconds down, no bounce, drive up. Unpaced reps measure your rhythm as much as your mechanics.
What the scorecard actually says
Depth against parallel is the headline — typical knee angle, and the share of reps that got there. Trunk angle in the hole catches a squat turning into a good-morning; it is a shoulder-to-hip body line, not a read of whether your back is rounding, because no spine landmark exists. Tempo and whether depth fades across the set tell you if the last reps are still the same lift. A first rep that is much shallower than the rest is a warm-up problem, not a depth problem, and the fix is different.
What to actually fix
Above parallel is the most common finding. Sit back and down between your feet rather than straight down, and squat to a box set at the height where your hip crease meets your knee. If the first rep is much shallower than the rest, do two or three genuinely easy sets before the one that counts. Torso too far forward is often an ankle restriction showing up further up the chain, and sometimes just a long femur relative to your torso, which is anatomy and not a fault — widen the stance slightly and turn the toes out before you go looking for a mobility fix. Inconsistent tempo: put a count on it. Three seconds down, no bounce, drive up. That also makes every future measurement more meaningful, because it removes rhythm as a variable. Do not solve a shallow squat with the heels planted and a mid-range shin by treating it as an ankle-mobility problem — the numbers usually say otherwise.
What we measure
• Depth against parallel, as a knee angle, and the share of reps that hit it
• Typical depth and whether it fades through the set
• Trunk angle in the hole — shoulder-to-hip line, with an explicit back-position caveat
• Stand / lockout between reps
• Down:up tempo on a controlled descent
What this camera cannot score
• Knee valgus — across 16 studies, 2D video vs true 3D valgus in a squat pooled to r = 0.127, not statistically significant
• Pelvic drop as a diagnosis — there is no published normal range for pelvic obliquity in a two-legged squat
• Butt wink — the pose model has no spine keypoints, so lumbar flexion is invisible at any camera angle
• Knees past toes as a graded fault — blocking forward knee travel raises hip torque roughly tenfold
• Foot pronation or a true heel-lift measure — heel and toe landmarks are blank on iPhone Vision
• Bar path, bar speed, or load — the camera never sees the bar
Questions
Do I need two clips?
No. Grade the side-on clip. A back-view clip would be a screen for symmetry, not a score — there is no published normal for pelvic drop in a two-legged squat, and 2D knee valgus is not a valid measure. We refuse both rather than sell them.
Why does camera height matter?
At the bottom of a squat your hips travel back and your knees travel forward. Filmed from above, that difference becomes fake depth and you look shallower than you are. Lens at hip and knee height — about two feet off the floor — is the shot that is actually measuring depth.
How many reps do you need?
Five. Fewer than that, medians are noise. The clip is rejected with that reason and does not use quota.
Can I upload a clip I already filmed?
Yes. Upload is on-device: this phone or browser reads the file. Nothing about the video is sent to our servers. Use the original file, not a copy that went through Messages or WhatsApp — those re-encode and lose the detail pose needs.
Keep your results
The free check scores one clip and forgets it. An account saves every analysis, writes the 4-week plan around your worst band, and shows whether the number actually moved when you re-film. Five analyses a month, free.