A bike fit you can check for yourself
Film your client on the trainer with a camera you already have. Within a few minutes you have the joint angles per crank revolution, adjustments in millimetres with their spread, and a report that can go in the file.
- 172.5
- mm of crank arm as the ruler, not an estimate from height
- 23
- revolutions to average over, from thirty seconds of footage
- 6
- adjustments worked through, in the order they belong in
- 0
- frames that leave your practice
What this is
A bike fit is usually a two-hour appointment at a studio with a measuring system worth several thousand pounds. That works, but it is a separate appointment in a separate place, and what the client takes home is a sheet with the final numbers. This is something else: a measurement you take yourself, in your own room, in a quarter of an hour, with a report that shows how every number came about.
The camera does not have to be special. What does matter: it sits on a tripod, square to the bike at bottom bracket height, and it runs at sixty frames per second or more. That last one is the one people forget, and it decides everything: at thirty frames per second there are only twenty-four frames in a revolution, and then you are measuring the camera instead of the rider.
The computing happens in the browser on your own machine. No footage goes to a server; what gets sent are coordinates. That is not only pleasant for data protection, it is the reason a practice is able to use this on recordings of people at all.
How it works
- 1
You film twenty to thirty seconds
Side view from the drive side, camera square at bottom bracket height, steady cadence. At ninety revolutions per minute that gives forty-five revolutions to average over: nearly twice what a running gait analysis has in stride cycles.
- 2
The crank arm is the ruler
You enter the crank arm length; it is usually stamped on the inside. The pedal spindle traces a circle with exactly that radius, and from that follows how many pixels make a metre. That is a real measurement, not an estimate from body height.
- 3
Every measurement is read at an exact crank position
Not "which frame looks most like bottom dead centre", but "what was the knee angle when the crank stood at exactly 180 degrees". That follows the six o’clock method from the international consensus statement, and it removes up to 4.5 degrees of crank angle worth of rounding error.
- 4
You get a report and an overlay video
The report holds the measured values with their spread, how they run across the revolution, and the fitted pedal circle so you can check with your own eyes that the scale is right. The overlay video shows where the numbers come from, and where the model struggled, because there the line goes faint.
What gets measured
Per crank revolution, then summarised as a median with the spread and the number of revolutions beside it. A mean across the whole clip hides one odd revolution; a median across twenty-three does not.
Knee and hip
- Knee flexion at bottom dead centre: the primary measure for saddle height, at the six o’clock position
- The same angle as an included angle, because fitting practice counts it the other way round
- Knee flexion at top dead centre and the total range of motion
- Hip angle at the top: how much room is left between thigh and torso
Position on the bike
- Torso angle relative to horizontal
- Shoulder angle: the classic reach measure
- Elbow flexion
- Vertical movement of the hip, in millimetres
Pedalling
- Cadence, from the crank angle and not from a sensor
- Ankle range of motion: the pedalling style
- Smoothness of the pedal stroke, where the frame rate allows it
- Left-right differences, where both legs can be tracked reliably
Where the measurement stops
On synthetic footage, where the value is known by construction, knee flexion at bottom dead centre lands within six tenths of a degree. On real footage of a time trial bike the scale came out at 440.6 pixels per metre with a shape deviation of the pedal path of 3.8 per cent: and that scale agreed independently with the wheelbase in shot.
What comes on top of that, and what no supplier should wave away: 2D video systematically underestimates joint angles compared with 3D motion capture. That error is systematic, so it falls away entirely as soon as you subtract two measurements from each other. Which makes before-and-after the strongest way to use this, and holding a single absolute number against an external norm the weakest.
The software says so when something is not right. If the camera sits at an angle the pedal path becomes an ellipse and the analysis stops. If the far leg is filled in by the model rather than seen -- on a bike the rule rather than the exception -- it measures the drive side only, and that goes in the report.
What is not in it: saddle width, shoe wedges and bar width. Those need a sit bone measurement or a physical examination and belong with the fitter, not with software that has only seen a video.
The footage stays on your own machine. Pose estimation runs in the browser; what goes to the server are coordinates. There is no route in the software that accepts a video file, and that is not an omission but the design.
What it costs
From € 28 an analysis on a bundle, or a subscription for unlimited use within your practice. Bundles do not expire. From around 6 analyses a month the subscription is cheaper, which we work out here instead of letting you discover it. If you also do running gait analyses, the combined plan is considerably cheaper than two separate subscriptions.
See the pricingCommon questions
- Does this replace a bike fit with a specialist?
- No, and it is not meant to. A fitter sees things no camera does: how someone sits on the saddle, where they feel pressure, how the hip moves when you put a hand on it. This is the measuring part of that work, and it makes that part repeatable and checkable. Saddle width, shoe wedges and cleat rotation belong with the fitter and are deliberately not in here.
- What camera do I need?
- A phone from the last few years will do, provided you set it to sixty frames per second: a setting that is often on thirty. Beyond that: a tripod, and as far back as you can get with the zoom in rather than close up on a wide angle. The latter distorts, and that comes back in every measurement.
- Does it work for time trial and triathlon bikes?
- Yes, and the discipline is the first thing you tell it. A road rider sits at around forty degrees of torso angle, a time triallist at around twenty. Letting the road profile loose on a time trial bike does not give slightly different guidance, it gives guidance pointing the wrong way. There are three profiles -- road, time trial and touring -- and you can add your own with your own target values.
- Do I get adjustment suggestions?
- If you ask for them. That is a separate choice and it is off by default, because the measurement is the product and the guidance is a recommendation with its own caveat. Turn it on and you get six adjustments worked through in the order they should be carried out: each adjustment changes the starting point of the next, and the chain accounts for that.
- How is saddle height arrived at?
- From knee flexion at bottom dead centre and the measured segment lengths of this person. Cycling uses rules of thumb like "a degree is three millimetres", but that only holds for the legs it was once established on. Here it is calculated: for an extended leg it comes out at one and a half millimetres per degree, for a deeply flexed one at a little over three.
- What if one leg cannot be tracked?
- Then it measures on the other side and says so in the report. On a bike the far leg passes behind the frame and behind the near leg, and a pose model then returns not a gap but a plausible guess. That is caught because both pedals sit on the same spindle and their circles therefore have to coincide. Measuring one side is not a fallback: a fit is done on the drive side.
- Is this a medical device?
- No. The intended purpose is comfort, efficiency and performance. It makes no diagnosis, treats nothing and says nothing about health. That boundary sits not only in the documentation but in the language of every text the software produces, and it is enforced by a test that checks all four languages.
Do you also do running gait analyses?
Then the combined plan is cheaper than two separate subscriptions. The same camera, the same way of working, the same kind of report: only a different sport.
Gait analysis alone
€ 129
a month
Bike fitting alone
€ 149
a month
Together
€ 199
a month: € 79,00 saved
This is a position measurement for comfort, efficiency and performance. It is not a medical examination and says nothing about complaints or injuries.