SMCAR transforms standard mobile device with iOS 17+ using iPhones or iPads into clinical-grade, markerless motion capture systems for remote rehabilitation. Unlike traditional methods requiring costly labs or failing to provide real-time analytics, SMCAR leverages ARKit and a hybrid CNN-Transformer pose estimation algorithm to dynamically track 22 body joints with <15° mean absolute error. It provides dual real-time feedback: a virtual avatar mirroring patient motion and live graphical displays of joint angle fluctuations. The application integrates a hybrid guidance system that overlays the user’s augmented reality skeleton alongside an instructor video, adjusting playback speed based on performance gaps. SMCAR automates active range of motion (AROM) measurements without clinician-administered goniometer tests, stores longitudinal performance data, and supports standardized motor impairment assessments (e.g., Fugl-Meyer). By delivering laboratory-grade kinematic analysis on consumer devices with <10ms rendering latency, SMCAR democratizes access to precision neurorehabilitation, enabling continuous remote monitoring and adaptive exercise progression.