How Does FreeMoCap Work Step by Step?

Updated

FreeMoCap turns video from two or more ordinary cameras into 3D motion data through four stages: set up cameras, calibrate the space, record synchronized footage, then process it locally into formats like CSV, FBX, and .blend. It is markerless, meaning you do not attach physical markers to the subject, and it is built to run on everyday devices such as webcams, GoPros, or smartphones rather than specialized motion capture hardware. The workflow below follows the four steps described on freemocap.org.

Step 1: Set Up Two or More Cameras

Position at least two standard cameras around your space. FreeMoCap's "universal design" is intended to produce high-quality results from common devices, so you are not required to buy dedicated mocap equipment.

Practical considerations for this stage:

  • Minimum of two cameras — the site specifies "two or more."
  • Surround the space — cameras should view the capture volume from different angles so the 3D reconstruction has enough viewpoints.
  • Everyday devices are supported — webcams, GoPros, and smartphones are all named as valid options.

Step 2: Calibrate the Space

Wave a printed ChArUco board within view of all cameras. This single action does two things:

  1. Synchronizes the camera feeds so frames line up across devices.
  2. Defines the 3D coordinate system needed for accurate spatial reconstruction and tracking.

Calibration is what makes multi-camera reconstruction possible — without a shared coordinate frame, the software cannot place a joint in 3D space consistently across views.

Step 3: Record the Motion

You have two recording paths:

Path How it works
Through FreeMoCap Record synchronized footage directly in the FreeMoCap interface while performing your movements naturally
Offline on mobile devices Record on your devices first, then import the video files for later processing

Both paths feed into the same processing stage, so choose based on whether you want live synchronization or the flexibility of recording away from the computer.

Step 4: Process and Reconstruct

The local processing engine converts your video into 3D data. Processing happens on your own machine rather than in the cloud, based on the site's description of a "local processing engine."

Output formats include:

  • CSV — for analysis
  • FBX — for professional animation pipelines
  • .blend — for Blender workflows
  • "many others" — the site notes additional formats beyond these

What You Get at the End

The result is full-body 3D movement data for humans, animals, robots, or other objects. FreeMoCap describes itself as connecting open-source computer vision and machine learning tools into one framework, so the output is meant to slot into existing research and animation pipelines rather than lock you into a proprietary format.

Who This Workflow Suits

The project names 3D animators, game designers, athletes, coaches, performers, scientists, engineers, clinicians, and doctors as intended communities. Its stated goal is a system that serves a professional research scientist while remaining intuitive to a 13-year-old with no technical training — so the four-step flow above is deliberately kept minimal.

If you want to try it, the site offers Download, Docs, and Code entry points. Start with two cameras and a printed ChArUco board, and expect the calibration step to be the one that most affects reconstruction accuracy.