Patient-specific muscle geometry is not only an interesting clinical tool to evaluate different pathologies and treatments, but also provides an essential input data to more realistic musculoskeletal models. The protocol set up in our study provided the 3D-patient-specific geometry of the 13 main muscles involved in the knee joint motion from a few selected magnetic resonance images (MRIs). The contours of the muscles were identified on five to seven MRI axial slices. A parametric-specific object was then constructed for each muscle and deformed to fit those contours. The 13 muscles were obtained within 1 h, with less than 5% volume error and 5 mm point-surface error (2RMS). From this geometry, muscle volumes and volumic fractions of asymptomatic and anterior cruciate ligament deficient subjects could easily be computed and compared to previous studies. This protocol provides an interesting precision/time trade-off to obtain patient-specific muscular geometry.