compareeHub

YOU CAN BUILD THE MOTION CAPTURE RIG BIG-BUDGET FILMS USE

Optical motion capture normally costs thousands per camera — Dennis built a sixteen-camera rig himself, each one under a tenth the price of a commercial unit.

by Dennis (Made By Dennis)

FULL CAD BOM FIRMWARE DOCS

Open-hardwareAI

Built withRaspberry Pi3D printing

difficulty
●●●●●
time
weeks, not a weekend
license
Apache-2.0
repo
repo ACTIVE384 stars

WHAT YOU’LL NEED

Videos
1
Jump to section

COMPAREE VERDICT

This is a full optical motion capture system — the kind studios use to turn an actor's movement into a 3D character — built from scratch around sixteen cameras Dennis designed himself. Each camera is a Raspberry Pi Compute Module on a custom circuit board with its own ring of synchronised infrared LED strobes; together the sixteen cameras fire pulses of up to 2.5 kilowatts. The sensor runs at 120 frames per second at full resolution, with 240 FPS designed in, trigger timing between cameras is better than 50 nanoseconds, the array processes over four billion pixels per second, and he reports tracking accuracy better than half a millimetre. He says each of his cameras costs less than a tenth of a comparable professional unit, which he puts at over 2,000 dollars. The files are published: the housing and wand parts on Printables, the boards, code and documentation in his GitHub monorepo. The single thing most likely to go wrong is underestimating the project before you start. This is not one camera and a weekend — it is sixteen cameras, sixteen boards, high-power LED drivers, a calibration routine and computer vision software that has to run in real time. If you have never designed a PCB or written camera synchronisation code, this will be a months-long learning project. If you have, it is still weeks of assembly and calibration. There is no bill of materials you can send to a supplier; you work from the KiCad boards and the docs. For someone with the skills and the workshop space, this is one of the most capable motion capture builds published under an open licence. For someone without those skills, it is the wrong first project — or you can buy his ready-made beta cameras.

GOOD TO KNOW

  • —CAD for the printable camera housing is on Printables under Creative Commons Attribution.
  • —KiCad PCB designs (CERN-OHL-P v2), code (Apache 2.0) and documentation are in a GitHub monorepo; both licences allow commercial use.
  • —No explicit bill of materials — you extract part numbers from the PCB designs and code.
  • —The monorepo contains dozens of other projects; motion capture sits inside /pkg/vision/mocap.
  • —Sixteen cameras means sixteen Raspberry Pi Compute Modules, sixteen custom boards, sixteen IR LED rings and a calibration wand.
  • —A second video covers the computer vision side: marker detection, matching across cameras and skeleton solving.

Parts to buy

6 items

From our check of the build. Exact quantities and part numbers are in the creator’s BOM.

  • Sixteen Raspberry Pi Compute Modules (CM5 or CM4)Find
  • M12 lensesFind
  • PoE+ network switch and Ethernet cablesFind
  • Reflective markersFind
  • Calibration wandFind
  • Host computer to run the capture softwareFind

BUILDS OF THE WEEK

Five open-source builds worth your weekend, every week.

Checked like this one: what’s really in the repo, what it costs, how hard it is. One email, unsubscribe anytime.

Can I build this?

PrintCamera housing for sixteen units, plus a calibration wand — files on Printables
BuySixteen Raspberry Pi Compute Modules (CM5 or CM4), sixteen sets of custom PCBs (compute, LED ring and AR0234 camera boards, fabricated from the KiCad files in the repo), M12 lenses, a PoE+ network switch and Ethernet cables, reflective markers, a calibration wand, and a host computer to run the capture software
ToolsPCB assembly capability (or a service that will build from Gerbers), 3D printer, soldering station, multimeter, oscilloscope for debugging synchronisation, large space for camera placement and calibration
SkillsAdvanced — PCB fabrication, embedded Linux, real-time computer vision, camera calibration, high-power LED driver design. If you have never built a circuit board or written threaded camera code, start with a simpler project.
TimeWeeks for someone experienced — board bring-up alone is days per camera, assembly is another week, calibration is iterative, and the software stack assumes you can read Rust and debug timing issues.
Cost$$$ — dominated by sixteen Compute Modules and PCB fabrication. Dennis claims under a tenth the cost of a comparable professional camera, but with no BOM you are pricing parts yourself.
SafetyHigh-power infrared LEDs (2.5 kW peak) — do not look directly at the strobes or run them outside their intended duty cycle. Power supply design for sixteen cameras requires proper current limiting and thermal management.

Build at your own risk. Projects involve tools, electronics and sometimes mains voltage — follow the creator’s safety notes.

Videos

Hardware build, camera design, synchronisation and tracking performance — the software side is in a second video on the same channel

More builds like this

All projects
This little box runs a voice-controlled AI agent fully offline on one Raspberry PiAITurn any glasses into AI smart glasses for under 25 dollarsAIBuild a talking AI assistant on a breadboard and host its brain yourselfAI

Gallery

Start here

Navigation into the creator’s own docs — we don’t rewrite the guide, we route you to the source.

  1. 1.Watch both videos (Hardware first, then the software walkthrough — you need both halves)
  2. 2.Read the motion capture documentation (The monorepo contains many projects — motion capture is under /pkg/vision/mocap, and its index page has a Getting Started guide and links to every hardware doc)
  3. 3.Download the camera housing STLs (CC BY licence — print one housing first to check tolerances before printing sixteen)
  4. 4.Export Gerbers from the KiCad boards and collect part numbers (No explicit BOM — you will be reading schematics and searching datasheets)

Resources

Documentation, files and community threads for this build — we link straight to the original sources and never rehost the creator’s files.

KNOWN ISSUES

  • Sixteen cameras means sixteen of everything — if you order parts for one and test, you are still ordering parts fifteen more times. Budget and storage space scale with camera count.
  • No bill of materials — you extract part numbers from the PCB designs and firmware. If you have never done this, add days to your timeline.
  • The monorepo is a personal workspace with dozens of projects. Motion capture is one subdirectory among many, and the build system assumes you know Rust and Bazel.
  • Synchronisation is the hard part. The cameras strobe within 50 nanoseconds of each other — if your PCB layout or firmware has timing bugs, markers will appear to jump between frames.
  • Calibration is iterative and requires a large open space. You move a wand through the capture volume while the software solves camera positions — if your space is too small or cluttered, calibration fails.
  • High-power infrared LEDs generate heat and need proper current limiting. If the driver design is wrong, you will burn LEDs or worse.

Can I build just one camera to test?

Yes, and you should — but motion capture requires triangulation, so you need at least two cameras to see 3D positions, and more cameras give better coverage and occlusion handling. One camera will prove your PCB works; it will not capture motion.

Do I need exactly sixteen cameras?

No — the design scales. Sixteen is what Dennis built for full-body capture with good occlusion resistance. Fewer cameras work if you accept smaller capture volume or more blind spots.

What is the actual cost per camera?

Dennis says less than a tenth of a comparable professional unit, which he puts at over 2,000 dollars — so under 200 dollars per camera by his estimate. You are pricing parts yourself, so verify that against current component costs.

Is there a pre-made PCB I can buy?

Not as a bare board kit — you export fabrication files from the KiCad projects in the repo and send them to a PCB service. But Dennis also sells complete, pre-focused cameras in a beta run, with a wand kit included; the repo documents exactly what ships.

What capture software does it use?

Custom software in the same monorepo — it finds markers, matches them across cameras and solves a skeleton in real time. The second video covers the computer vision side.

Community builds

No community builds yet — be the first, we feature the best ones.

Discussion1

FROM THE COMPAREE TEAM

Sixteen cameras, triggered to within 50 nanoseconds of each other, and the author reports sub-millimetre tracking accuracy. What would you capture first if you built this?

CompareeTEAM1mo agoedited

Practical notes from our verification: the GitHub repo is a monorepo with dozens of projects, and motion capture lives in one subdirectory (pkg/vision/mocap), so expect to navigate. The code is Apache 2.0 and the board designs are under CERN OHL v2 Permissive; the Printables page carries the 3D-printed housing parts. Documentation is more complete than you might expect: a getting-started guide, pages for each camera board, the OS image, the microcontroller firmware and lens selection, and pre-built host software for Linux and Windows. There is no single shopping-list BOM, so you work from the KiCad boards and docs to source parts. This is still not a weekend kit build: the rig in the video uses sixteen custom cameras, each with its own Raspberry Pi compute module, AR0234 sensor board and high-power IR LED ring, plus a PoE+ switch, a calibration wand and a room to calibrate in. The author also offers ready-made cameras if you would rather not fabricate boards. If you have the skills, it is one of the most capable open motion capture systems we have seen. Correction (4 October 2026): we re-checked this page line by line against the project's own repository, documentation and videos, and fixed errors in earlier versions.

Dennis (Made By Dennis)

Dennis builds ambitious hardware from scratch and documents it on his Made By Dennis channel — his own 3D printer with custom firmware, a 1 petabyte server, and this motion capture system. The mocap rig came from wanting studio-grade tracking without paying over 2,000 dollars for every entry-level professional camera.

GitHub

Star the project on GitHub

DISCLAIMER

  • Comparee is not the author of the projects featured here. All rights to each project belong to its creator — every page links to the original source, and we never host creators’ files.
  • Information is provided without warranty and may become outdated as projects evolve. Prices are indicative bands only — always check the creator’s parts list for current costs.
  • Building and operating any project is at your own responsibility. Protective equipment, safe workshop practice and compliance with local regulations are the builder’s responsibility.
GitHub repo