MovementModeler のアプリ分析 12月9日

MovementModeler converts your videos into clean, accurate stick-figure motion clips using fast, fully on-device processing. The app is designed for creators, athletes, coaches, researchers, and computer-vision professionals who want a simple and reliable way to extract, analyze, and export human motion—without accounts, cloud services, or privacy concerns. Load any video from your library and MovementModeler automatically generates a stick-figure preview for each detected person. The app supports multi-person scenes through automatic track identification, maintaining consistent labels even during movement, occlusion, or crossing paths. Each person’s motion track can be isolated in any combination for both preview and export, allowing you to focus on a single subject or compare individuals separately. Refine your results with real-time controls designed to clean up motion artifacts. Adjust filtering to reduce jitter, stabilize joint movement, fill missing limbs, and smooth noisy frames. Playback speed automatically matches your source video, and frame-by-frame scrubbing gives you precise control for detailed review. When you’re ready to export, choose between a clean MP4 stick-figure clip or BODY-25 JSON keypoint data—ideal for research, robotics, sports analysis, rehabilitation, animation, and machine-learning workflows. Each exported JSON file contains per-frame keypoints, confidence values, and consistent track assignments for multi-person videos.