核心要点 (BLUF)： Marker-less motion capture realtime, online in your browser: track your body pose and hands using a standard webcam. Everything runs locally on your device without sending video to any server or charging tokens.
The MediaPipe PoseLandmarker neural network runs directly inside your browser via WebAssembly and WebGL/WebGPU. It continuously processes webcam video frames and detects 3D coordinates of 33 body joints in real time.
No specialized suits, optical markers, or depth sensors are needed: an ordinary laptop or USB webcam with decent lighting is all you need.
Your movements are mirrored instantly onto a 3D mannequin in the interactive viewport. You can freely orbit, pan, and zoom around the 3D scene using your mouse or touch gestures.
Record motion takes and export them to JSON or BVH formats, ready for import into Blender, Unreal Engine, Unity, or other 3D animation suites.
Video processing is 100% on-device: camera frames never leave your computer and are never stored or transmitted to our servers.
The tool is completely free, does not require a subscription, and charges zero tokens.
Yes. The engine executes locally in your browser and charges zero tokens.
No. All neural network inference takes place locally on your device; video never leaves your browser.
Any standard laptop webcam or USB webcam at 720p 30 FPS or higher under good lighting works well.
Click "Record Motion", perform your moves, stop recording, and download the motion as JSON or BVH.