FreeMoCap vs PoseCap

FreeMoCap records and reconstructs full-body movement as 3D motion-capture data. It is free, open-source software for researchers, animators, athletes, and creators. Capture can use two or more standard webcams, GoPros, or smartphones; a printed ChArUco board synchronizes camera feeds and defines the 3D coordinate system. Users can also record video offline on mobile devices and import it later for processing. Results can be exported in formats including CSV, FBX, and .blend files. A Blender add-on loads processed recordings into Blender as a rigged skeleton. Processing happens locally, and the homepage says video data does not leave the user's machine. A backend server download supports headless machines, remote capture rigs, and API use. The download page identifies the current release as alpha. Optional anonymous usage pings may include IP address, timestamp, and camera information, but users can disable them in the GUI. The project directs questions to Discord and bug reports or feature requests to GitHub. Its post-hoc guide notes that RTMPose produces 3D data but does not currently have working Blender export.

PoseCap is a free Blender extension for streaming markerless, full-body poses from a webcam or video into a character in real time. It can stream live motion, record movement as Blender timeline keyframes, capture a single pose, and manage SMPL-X models. A separate engine process sends pose data to Blender over a local connection. MediaPipe Lite provides CPU-first capture without a GPU or provider account, while the PEAR backend provides higher-fidelity capture and requires an NVIDIA GPU with driver R570 or newer. The getting-started guide describes importing a Mixamo or Unreal character and converting it in one click. Blender 4.2 LTS or newer is required. PoseCap lists Windows 10/11 as its current target and provides a native Linux installation path; macOS is not currently planned. SMPL, SMPL-X, and FLAME models are separate licenses. Poses are pelvis-locked because monocular depth estimation cannot generally recover world translation.