Blender Free Plugins · Tutorial bf-07

Webcam Motion Capture with BlendArMocap

Capture full-body, hand, and face motion from a regular webcam using the free GPL BlendArMocap addon, then transfer it to your Rigify rig. Part of the MasterAllArts Creative 3D Roadmap.

Part of the MasterAllArts Creative 3D Roadmap .

Plugin: BlendArMocap ↗ · Free · GPL-3.0 · advanced

BlendArMocap by Denis Sutter (cgtinker) is a free, GPL-3 addon that turns any webcam into a markerless motion-capture rig. It runs MediaPipe inside Blender, captures hand, body, and face landmarks per frame, and bakes them into an empties-driven control rig you can transfer onto your Rigify character.

Quality is not Rokoko-tier, but for a free webcam mocap workflow it’s the best option in the Blender ecosystem today.

What you need

  • A webcam (or a phone connected as a webcam).
  • Python wheels for MediaPipe and OpenCV — BlendArMocap can install these for you, but Python version compatibility matters. See “Install” below.
  • Reasonable lighting and a contrasting background. The model degrades fast in poor light.

Install

  1. Download the latest release ZIP from the BlendArMocap releases page. Match the release to your Blender version — the addon ships separate builds per Blender major version because of bundled Python wheels.
  2. Edit → Preferences → Add-ons → Install and select the ZIP.
  3. Tick the checkbox to enable.
  4. Open the addon’s preferences (expand the row in the addon list). Click Install Dependencies. This pip-installs mediapipe, opencv-python, and protobuf into Blender’s Python.
  5. Restart Blender. This is required for the new wheels to load.

If “Install Dependencies” fails, your network blocked pip or Blender’s Python doesn’t match the wheel ABI. The maintainer documents per-version troubleshooting in the README — start there.

Capture motion

After restart, BlendArMocap appears in the N-panel under a tab called BlendAR.

  1. In the BlendAR panel, pick a detection type:
    • Pose — full body
    • Hand — hands only (for sign or precision)
    • Face — facial landmarks
    • Holistic — all three at once (slowest, most accurate)
  2. Pick Webcam as the input source. (You can also feed a video file, which is often more reliable than a live capture if you want to retake.)
  3. Click Detect.

A viewport window appears showing the webcam feed with landmark overlays. Move around for as long as you want to capture. Press Q in the webcam window to stop.

The addon writes empties into your scene — one per landmark — and animates them with location keys for every detected frame.

Generate the control rig

Captured empties are not yet usable on a character. Convert them to a transferable rig:

  1. Select the captured empties (BlendArMocap parents them under a single armature; pick that).
  2. In the BlendAR panel, click Transfer → Generate Animation.

This creates a placeholder armature (the “BlendAR rig”) whose bones are constrained to the empties. It’s the source rig for retargeting.

Transfer onto your Rigify character

You have two paths from here:

Path A — BlendAR’s built-in transfer. The addon ships with a Rigify preset:

  1. Select the BlendAR rig and the target Rigify rig.
  2. Transfer → Bake to Rig.
  3. Pick “Rigify” as the target preset.
  4. Bake the action.

Path B — Manual retarget. Treat the BlendAR rig like a Mixamo source and use the Retarget extension or Expy Kit. The bone names are documented in the BlendArMocap repo.

Path A is faster. Path B gives you cleaner control if you’ve already invested in a preset library.

Cleanup

Webcam mocap output is always noisy. Expect to:

  • Smooth the action in the Graph Editor with Key → Smooth Keys or a Smooth modifier on the F-curves.
  • Trim the first and last few frames where the detector was warming up.
  • Manually fix foot-slide. There’s no foot-IK constraint coming from BlendArMocap.
  • For face capture, blend the captured face action at low influence over a hand-keyed performance.

Treat BlendArMocap output as a rough first pass, not a finished take.

What it can and can’t do

Good at:

  • Single performer in clear light, full body visible.
  • Captured hand poses for reference.
  • Face capture for placeholder lip-sync and broad expressions.

Bad at:

  • Two performers (only one tracked at a time).
  • Heavy occlusion (the limb disappears mid-take).
  • Anything requiring inertial accuracy — feet won’t plant cleanly.

Common pitfalls

  • Dependencies install reports success but Blender errors on Detect. Almost always a Python ABI mismatch. Verify Blender’s Python version (Help → Save System Info) matches the BlendArMocap release you installed.
  • Captured empties at world origin only. The webcam feed never resolved a person — check lighting and that nothing else is using the camera (Zoom, Discord).
  • Action looks correct on the BlendAR rig but garbage on Rigify. Rest-pose mismatch — see the rest-pose section in bf-04.

You’ve completed the track

You now have the full free-Blender retargeting workflow:

  • Generate a character (MB-Lab)
  • Rig it (Rigify)
  • Get animations onto it (Mixamo, your own webcam mocap, or both)
  • Transfer those animations across rigs (Retarget extension, NLA constraints, or Expy Kit presets)

If you want to push further, the next step is paid territory — Auto-Rig Pro for production-grade auto-rigging and remap, or Rokoko Smartsuit + their free Blender plugin for actual sensor-based mocap. But for free, you’ve now seen the full toolkit.