Objective This research is aim to develop a modeling method based on motion data, which can improve the static and dynamic matching precision and the reliability of kinematic and kinetic calculation. Methods Model nonlinear scale is implemented by position adjustment, interpolation function calculation and scaled model generation. For the verification of this method, two open dataset (GC3 and GC5) are used to build the nonlinear scaled models and calculate the limb lengths and joint reaction forces. The results are compared with those calculated by Anatomical Landmark Scale and linear scale method. Results The maximum discrepancies between limb length of nonlinear scaled model and actual model are 14.74 mm, which are in the region 4.0±13.8 mm reported by other research. Marker errors of scale and inverse kinematic calculation can fulfill the requirement of OpenSim. As for calculated joint reaction forces, the RMSEs (GC3: 0.40 BW, GC5: 0.34 BW) are less than those of Anatomical Landmark Scale (GC3: 0.64 BW) and OpenSim linear scale method (GC5: 0.40 BW). Besides, the results of Monte Carlo analysis indicate that, with the variation of initial positions of model markers, the range of joint reaction forces errors are less and limb lengths fluctuate within 5%. Conclusions This nonlinear scale method is effective and in current verification condition, it can improve the efficiency of modeling process and raise the precision of simulation results.