We propose a computer vision model to assess the risk of sarcopenia in functional movement video clips. Sarcopenia progressively reduces muscle mass and strength with age, posing a significant threat to the well-being of seniors. Early detection and timely intervention can significantly improve an individual’s life and alleviate pressure on the healthcare system. Our model includes a 3D posture keypoint detector and a transformer classifier. The posture keypoint detector identifies 16 keypoints that form 4-Vector and 7-Vector input configurations capable of distinguishing individuals with sarcopenia from those without. However, the differences in these configurations between individuals with sarcopenia and those without are too subtle for human observation. Therefore, we trained the transformer classifier to assess the probability of sarcopenia risk in video clips featuring five specific functional movements. We verified our approach through experiments involving 20 sarcopenia patients and 20 individuals without sarcopenia.