AI Running Coach Analyzes Your Form From Plain Video

A team of researchers in Spain has built a system that can analyze a runner's form using nothing but a regular video. The tool, called VideoRun2D, uses AI body-tracking technology to follow a runner's movement on screen and calculate the angles of their hips and knees as they sprint. In tests with 44 professional runners and 314 sprints, it matched expert manual measurements within about 5 to 9 degrees. That's roughly the width of two fingers held at arm's length — close enough to spot meaningful flaws in stride.

Today, proper biomechanical analysis usually costs thousands of dollars. It requires special motion-capture suits, multiple cameras, or pressure-tracked treadmills. VideoRun2D removes that barrier. A coach could film a sprinter with a smartphone, upload the video, and get instant feedback on whether their hip extends far enough or if their knee collapses inward. That's useful not only for elite athletes but also for weekend joggers trying to avoid knee pain or shin splints.

The system isn't flawless yet. When runners speed up, the AI's estimates get less accurate, and it occasionally misreads a limb against a busy background. But the researchers added a clever cleanup step that detects weird data points and smooths them out, cutting the error noticeably. They hope to expand beyond sprinting to longer runs and other sports.

What makes this exciting is accessibility. The underlying AI models are open-source and the demo runs on standard computers. In a few years, a similar tool could live inside fitness apps, smartwatches, or gym cameras. For everyday runners, it could mean a personal biomechanics expert in your pocket, offering simple advice like "shorten your stride" before an injury happens.