Biosystems One Inc.

Measure movementfrom ordinary video.

MotionLab turns a phone clip into joint angles, velocities and publication-ready figures — with the calibration, filtering and provenance a reviewer will ask about. It runs entirely in your browser, and your footage never leaves your computer.

No install No upload Free to learn on
Frame 0 Frame 24

Where it is used

Gait and clinical movementJoint angles through the stride, normalised to percent of cycle so trials of different length compare directly.
Occupational biomechanicsTask analysis, posture and reach envelopes, and repeated-motion exposure measured from footage taken on the floor.
Sport and techniqueSegment velocities and joint kinematics from high-speed phone footage, with the true capture rate applied.
Rehabilitation and return to workRange of motion and symmetry tracked session to session, saved as projects you can reopen and compare.
Teaching and student labsA whole cohort can run real analyses on their own laptops without a licence server or a lab booking.
Equipment and product testingQuantify how a device, brace or tool changes movement, and export the figures straight into the test report.

What it does

Built for measurements you have to defend.

Every number MotionLab reports carries its provenance — the frame rate it was derived from, the calibration behind it, the filter applied, and whether a point was measured or estimated.

Frame-accurate playback

Step one frame at a time with a scrubber that shows every frame as a tick. Shot in slow motion? Set the true capture rate so a 240 fps clip doesn't report speeds eight times too low.

Calibration that survives an angle

Draw across a known distance, or click the four corners of a floor tile. Plane calibration solves the perspective, so the camera doesn't have to be square on to the movement.

Body models, not blank canvases

Pick a side or front view and MotionLab builds the markers, the linked segment skeleton, and the joint angles that are conventionally reported — knee, hip, ankle, trunk lean — in one step.

Place each joint once

Click each joint on a single frame and tracking starts by itself, following every marker together. A 900-frame trial stops being 900 clicks per joint.

Estimates through occlusion

When a joint disappears behind the body, segment lengths and its visible neighbours place it geometrically. If it never reappears, the guesses are deleted rather than kept — and every estimate is marked as one.

Filtering you can cite

Zero-lag dual-pass Butterworth, the biomechanics standard. Choose your own cutoff, or let residual analysis derive one from your data and show you the curve it used.

Angles, distances, events

Mark foot strike and toe-off, then normalise the whole trial to percent of cycle — the convention for gait — so trials of different length compare directly.

Output that passes review

Publication figures on white at print resolution, CSV with a provenance column per marker, and a report whose methods table flags an unconfirmed frame rate or a default calibration in red.

Methods and rigour

Every number arrives with its working shown.

MotionLab records how each result was produced and puts it in the export. This is the methods table from a real report — you can paste it into a manuscript, hand it to a reviewer, or file it with a test record.

gait_trial_02 — report.pdf

Methods and provenance

Source filegait_trial_02.mp4
Resolution1920 × 1080 px
Frame rate240 fps at capture; file plays at 30 fps. All times use the capture rate.
Spatial calibrationPlane, four-point homography, 1.200 × 0.600 m reference
Coordinate conventionOrigin at reference corner 1, Y positive upward
SmoothingZero-lag Butterworth, 2nd order, dual pass, 6.5 Hz cutoff
Cutoff selectionResidual analysis
DifferentiationCentral finite difference
Analysed rangeframes 120–480 (1.500 s)
Time originframe 148 — right foot contact
Point provenance9 placed by hand, 3 204 tracked, 47 estimated from segment lengths
Segment model8 segments: Ear–Shoulder, Shoulder–Hip, Hip–Knee, Knee–Ankle…
Anything unconfirmed is flagged in red rather than hidden — an assumed frame rate, a default calibration reference, or points that were estimated rather than measured.

The methods behind it

Nothing here is a proprietary black box. Each step is a published procedure you can look up, argue with, and name in a paper.

  • FilteringZero-lag dual-pass Butterworth with the cutoff correction from Winter, Biomechanics and Motor Control of Human Movement — the standard in the field, applied forward and backward so no phase shift is introduced.
  • Cutoff selectionResidual analysis after Winter, or residual autocorrelation after Challis (1999). MotionLab plots the curve it used, so the choice is visible rather than asserted.
  • Perspective correctionA planar homography solved from four known points, so a camera that is not square on to the movement still measures correctly.
  • TrackingZero-mean normalised cross-correlation with sub-pixel refinement of the correlation peak, and constant-velocity prediction between frames.
  • Gap handlingOccluded joints are placed from segment-length constraints using their visible neighbours. Estimates that are never confirmed by a re-acquisition are discarded, not kept.

Single-camera 2D cannot measure motion toward or away from the lens. MotionLab states that limit rather than working around it, and records the calibrated plane every measurement depends on.

The workflow

Five steps from clip to figure.

These are the same five numbered tabs you will see down the left of the app. The guide inside it tracks the same sequence and tells you exactly what is outstanding at each stage.

Step 1

Video

Open a clip and confirm its frame rate. Shot in slow motion? Set the rate it was recorded at, not the one the file plays back.

Step 2

Scale

Calibrate against a known distance, or the four corners of a known rectangle if the camera is not square on.

Step 3

Track

Choose a body model, click each joint once, and tracking runs on its own from there.

Step 4

Measure

Add the joint angles and distances you need. A body model creates the usual ones for you.

Step 5

Analyse

Graph anything against time or percent of cycle, then export the figure, the data, or the full report.

Your data

The footage stays yours.

Movement video of a real person is identifiable, often clinical, and frequently covered by an ethics approval. MotionLab is built so that it never has to leave the machine it was recorded on.

Nothing is uploaded by default Your browser reads the file straight from disk and decodes it in memory. There is no upload step and no server waiting for it.
The analysis runs on your machine Tracking, filtering, angles and graphs are all computed in the browser tab — not on our hardware.
We store JSON, and nothing else A project is calibration, markers, coordinates and measurements — a small text file. There is no video or image storage on our servers at all, on any plan.
You decide what leaves Data moves only when you export a CSV, save a figure, or switch on syncing for a project.

Pricing

Free to learn on. Paid to publish with.

The whole analysis is free while you are learning the tool, teaching with it, or deciding whether it fits your lab. The moment the work leaves your hands — a figure in a paper, a report in a patient file, anything you are paid for — it is a Pro licence.

Study
$0

Learning, teaching and evaluation. No account needed to start.

  • The complete analysis — nothing is crippled
  • Calibration, body models, auto-tracking
  • Angles, distances, events, graphs
  • Save projects on your device
  • 100 MB workspace with a free account
  • No CSV, figure or report export

For coursework, self-teaching, classroom demonstration and trying MotionLab on your own data before you buy.

Start learning

Prices in USD. Annual plans are billed once and can be invoiced for institutions. Students and single researchers qualify for Pro at half price — just ask. The Study tier is licensed for learning, teaching and evaluation only.

Questions

Before you start.

Do I need to install anything?

No. MotionLab is a web page. Open it in Chrome, Edge, Firefox or Safari and start working. There is no plugin, no runtime, and nothing to get approved by an IT department.

What video formats work?

Anything your browser can decode — MP4 (H.264) and WebM are the safest. Some MOV, HEVC and ProRes files will not open; re-encoding to H.264 MP4 fixes it. For still work, JPG, PNG and WebP are all supported.

How accurate is it really?

Accuracy is bounded by three things: how carefully you calibrate, how well the camera is placed, and whether the movement stays in the calibrated plane. Single-camera 2D cannot measure motion toward or away from the camera — that is a property of the method, not of this software. MotionLab is explicit about the assumptions it is making, and records them in every export.

Can I use it for a publication?

Yes — that is what the report is for. It records the frame rate and where it came from, the calibration method and reference, the filter type and cutoff, the analysed range, and how many points were measured versus estimated. Anything unconfirmed is flagged rather than hidden.

Where exactly is the line between Study and Pro?

It is about the output, not about who you are. Learning the tool, teaching a class with it, and evaluating it on your own data are all Study. As soon as a measurement is going into a paper, a grant report, a patient record, or work somebody is paying you for, it needs a Pro licence. If you are a student and the answer is genuinely still unclear, write to us — we would rather say yes than have you guess.

Can I generate a report on the free tier?

No. Reports, figures and CSV export are Pro features on every plan and every tier, because the report is the thing you hand to someone else. Everything upstream of it — tracking, measurement, graphs on screen — is free.

What happens to my work if I stop paying?

Nothing is locked away. Analysis, tracking and saving projects stay available, so your existing projects keep opening and keep working. You lose the export formats, not your data.

Do you ever store my video?

No. Not by default, not as an option, and not on a paid plan. Your browser reads the file from disk and decodes it in memory; there is no upload endpoint to send it to. What a cloud workspace stores is a JSON file of your analysis — calibration, marker coordinates, measurements and settings — plus your account email. That is the entire list.

How do I sign in?

By email or with Google. Email sign-in sends you a verification link — there is no password to choose, forget, or have stolen from us, because we never store one.

Who is behind it?

Biosystems One Inc., a research and development company working across biomedical and biosystems engineering. MotionLab came out of biomechanics research needs that existing tools met awkwardly or expensively.

Open a clip and see.

No account, no upload, no trial clock. Learn on it for as long as you like — pay only when the work leaves your hands.