Main topics: Technical developments in movement science, Experimental studies in human movement science Introduction and aim: Quality of the popular Plug-in Gait (PiG) model [1] outputs depends primarily on marker positioning. Among them, thewandmarkersplacedon the thighsandshanksare especially difficult to position because they involve non-palpable anatomical points: the hip and the knee joint centers, respectively. This communication presents an optimal method for positioningwandmarkers using a real-time process.We examine the effect of thismethod on two operators (expert and non-expert). This process is expected to reduce the inter-session variability [2] and the maximum of the varus-valgus knee angle during swing phase. Patients/materials and methods: Gait data were collected on ten healthy subjects (62±10kg, 1.69±0.09m) with an optoelectronic motion capture system (Vicon, Oxford Metrics). Except the wand markers, all the PiG markers, plus medial markers were placed by an expert operator (5-year experience in gait analysis). A custom program using the Vicon RealTime SDK was implemented in Matlab. It indicated when the thigh wand marker, for example, belonged to the plan formed by the hip joint center, the lateral and medial condyles, with a tolerance of ±1mm. Firstly, each subject participated in four gait sessions, two with and twowithout real-time positioning by the expert. Between each session, the wand markers were fully removed. Four other gait sessionswere then carried outwith non-experts as operator (physiotherapist students without experience in gait laboratory). Kinematics PiG data were calculated by the use of the Biomechanical toolkit [3]. A two-way mixed-anova examined the effect of the realtime process and operators. Results: Real-time positioning significantly reduced the intersession variability of knee angles for the non-expert (Fig. 1). For the expert, only the flexion-extension was different. Similarly, a significant difference of 6 degrees (Table 1) was observed for the maximal knee var/valgus angle provided from the non-expert positioning. Thanks to the real-time process, themaximal var/valgus of the non-expert did not differ from the expert one.