Motion capture exporter
Turn an ordinary video of a person into animation data you can import into Blender, Unreal, or Unity. No suit, no markers, no rig — one camera. Runs entirely in your browser: the video is never uploaded.
About this motion capture exporter
Professional motion capture means a suit, a room full of cameras, and a five figure budget. This does the same job approximately, from one ordinary video, by estimating the position of each joint in 3D for every frame and converting those positions into the skeletal rotations a BVH file stores.
The output imports directly into Blender, Unreal, and Unity. The skeleton is built from your own subject’s proportions rather than a generic rig, so the motion sits correctly on the person who performed it; retarget from there onto whatever character you like.
Your video is never uploaded. Tracking runs inside your browser tab. Be realistic about accuracy: a single camera cannot see what it cannot see, so limbs that pass behind the body are estimated rather than measured. It is a strong starting point to clean up, not a replacement for a capture stage.
Common questions
- How do I import the BVH into Blender?
- File → Import → Motion Capture (.bvh). The armature arrives with the animation already applied. Use Blender’s retargeting tools if you want to drive a different character with it.
- How accurate is single-camera motion capture?
- Good for broad body motion — walking, gesturing, sport movements filmed side-on. Weaker on depth, on limbs that pass behind the torso, and on fine detail. There are no fingers or facial capture. Expect to clean up the curves.
- What kind of video works best?
- The whole body in frame for the entire clip, even lighting, a background that contrasts with the subject, and a camera that does not move. Film perpendicular to the main direction of movement.
- How long a clip can I export?
- Up to a few hundred frames — roughly 30 seconds at 24fps. Every frame is a model pass held in memory. Trim to the take you actually want rather than processing a whole session.
- Are hands and fingers included?
- No. The skeleton has fifteen joints covering the torso, head, arms, and legs — only what the pose model can genuinely observe. Adding finger joints it never saw would produce confident-looking invented data.
- Why are some frames missing from the animation?
- Frames where no person was detected are skipped rather than filled with the previous pose. A held pose would read as the subject standing still, which misrepresents what happened. The result panel tells you how many were skipped.