This paper addresses the problem of synthesizing plausible bipedal locomotion according to 3D anatomical reconstruction and general hypotheses on human motion control strategies. In a previous paper [Nicolas, G., Multon, F., Berillon, G., Marchal, F., 2007. From bone to plausible bipedal locomotion using inverse kinematics. Journal of Biomechanics 40 (5) 1048-1057], we have validated a method based on using inverse kinematics to obtain plausible lower-limb motions knowing the trajectory of the ankle. In this paper, we propose a more general approach that also involves computing a plausible trajectory of the ankles for a given skeleton. The inputs are the anatomical descriptions of the bipedal species, imposed footprints and a rest posture. This process is based on optimizing a reference ankle trajectory until a set of criteria is minimized. This optimization loop is based on the assumption that a plausible motion is supposed to have little internal mechanical work and should be as less jerky as possible. For each tested ankle trajectory, inverse kinematics is used to compute a lower-body motion that enables us to compute the resulting mechanical work and jerk. This method was tested on a set of modern humans (male and female, with various anthropometric properties). We show that the results obtained with this method are close to experimental data for most of the subjects. We also demonstrate that the method is not sensitive to the choice of the reference ankle trajectory; any ankle trajectory leads to very similar result. We finally apply the method to a skeleton of Pan paniscus (Bonobo), and compare the resulting motion to those described by zoologists.