We present a computer tool for testing walk hypotheses for human beings. This tool aims to generate plausible walking movements according to anatomical knowledge. To this end, we introduce an interpolation method based, on one hand, on morphological data and, on the other hand, on stance hypotheses and on footprint hypotheses. We want to test these hypotheses for application to the reconstruction of early hominid walking. We interpolate from a specific representation of the movement—a characteristic relative displacement. First, we use a motion capture system to acquire real movements of a walk cycle, and we propose to represent them by using a generic parametric model. Thus, we create a database of movements. The interpolation process produces, thanks to this database, a retargeted motion adapted to the morphology of the considered targeted skeleton. The interpolation is done according to three main hypotheses. The first concerns the reference stance, the second the lateral spacing between the feet, and the third the length of the step. In the introduction, we refer to related work. Then we propose the two following points of our method: the 3D representation of our motion representation and the multidimensional interpolation method applied to this representation. The interpolation method addresses morphological adaptation, and the use of an inverse kinematics solver addresses the computation of skeleton movements. The self-coherent validation process aims to test the coherence of the proposed method. The results propose an application to a virtual skeleton of Lucy (Australopithecus afarensis A.L. 288-1) reconstructed from real data. Finally, the relevance of the method for anthropological investigations and for animation purposes is discussed and future work is discussed with respect to the limitations of the proposed method.