Bike Fit Analyzer Open the analyzer →

Home / Guides / Filming

How to film a bike fit video that actually measures accurately

30 July 2026·6 min read

Almost every disappointing result from video fit analysis is a footage problem, not a fit problem. The pose model is doing fine — it is being asked to measure a leg that is partly hidden, or a rider filmed from three-quarters on. Get the camera right and the numbers become boring and repeatable, which is exactly what you want.

The short version

• Camera square to your side, not in front, not behind, not at an angle.
• Camera at hip height, roughly level with the bottom bracket to hip range.
• 3–5 m away if you can, so the lens is not distorting the near leg.
• 20–30 seconds of steady pedalling, in a gear you would actually ride.
• Light on you, not behind you.

Why square-on matters more than anything else

Every angle in a side-on fit read is measured in one plane. Your knee angle is the angle between your thigh and your shin as seen from directly beside you. Rotate the camera 20° toward the front of the bike and you are no longer measuring that angle — you are measuring its projection onto a tilted plane, which is always smaller than the real thing.

This is not a small effect. A camera well off axis can shift a knee reading by more than the entire width of the target band, which means a correctly-set saddle can read as too high. The analyzer tries to catch this: it compares how far apart your two shoulders and two hips appear horizontally against the length of your torso. Filmed from the side, your left and right shoulders almost overlap, so that spread is small. Filmed off-axis, it widens, and you get an off-axis warning on the result.

The practical fix is to line the camera up with your hip, not with the bike. Stand the phone where a line from the lens to your hip would be perpendicular to the bike's centreline.

Camera height, and the mistake almost everyone makes

Phones end up on the floor because the floor is where you can lean them. Filming upward from ankle height introduces exactly the same projection error as filming off-axis, just in the vertical plane — and it hits the torso and shoulder angles hardest, because those are measured against horizontal.

Hip height is the target. A chair, a windowsill, a stack of books, or a tripod all work. It does not have to be exact; being within about 20 cm of hip height is plenty, and much better than being 80 cm below it.

Distance and lens

Wide-angle phone lenses stretch whatever is nearest the edge of the frame. If you film from a metre away, your near leg sits in the distorted region for part of every revolution. Backing off to 3–5 m and letting the rider fill less of the frame trades a little resolution for a lot of geometric honesty. If your phone has a 1× and a 0.5× lens, use 1×.

What to do while you are filming

Pedal the way you actually ride. A steady, moderate cadence in a gear with some resistance gives the most consistent read. Two things to avoid:

• Coasting. Frames where you are not pedalling still get measured. A clip that is half coasting gives the analyzer fewer real strokes to work with.
• Performing. Sitting up unusually straight because you are on camera measures a position you never ride in. Look ahead, not at the phone.

Twenty to thirty seconds is the sweet spot. Below about ten seconds there are not enough pedal strokes to take a stable median, and the result starts depending on which frames happened to be sharp. Much beyond a minute just makes the analysis slower without making it better.

Trainer or outdoors?

A turbo trainer is easier and usually more consistent: the bike does not move, the camera does not have to track you, and you can film as long as you like. The one caveat is that people often sit slightly differently on a trainer than on the road — a little more upright, a little further back. If your trainer position and your road position genuinely differ, measure the one you care about.

Outdoors works if someone can hold the camera still and pan smoothly while you ride past at a steady speed, or if you can ride a fixed line past a mounted phone. Expect to need a couple of attempts.

Clothing and background

The pose model needs to find your joints, so it helps to make them findable. Fitted clothing reads better than loose — baggy shorts hide where the knee actually is, and a flapping jersey makes the hip ambiguous. Bare legs or tights both work fine.

For background, contrast is what matters. A rider in black kit against a dark garage door is hard for any vision model. Light on the rider and a background that is not the same colour as the rider solves most tracking failures. Backlighting — a bright window behind you — is the single worst case, because it silhouettes you.

The four mistakes, ranked by how much they cost

How to know your clip was good enough

The analyzer tells you. It reports how many pedal strokes it found, warns you if the footage looks off-axis, and greys out any angle where a joint was not confidently visible rather than guessing at it. A grey reading is not a failure of your fit — it means that joint was not measurable in that clip, and it is worth re-filming before you read anything into it.

The strongest signal is repeatability. Film two clips back to back without touching the bike. If the knee readings agree within a couple of degrees, your setup is giving you trustworthy numbers. If they disagree by ten, fix the filming before you touch the saddle.