Interactive systems, games, VR and multimedia systems require more and more flexible Virtual Humans with individualities. There are mainly two approaches: 1) Recording the motion using motion capture systems, then to try to alterate such a motion to create this individuality. This process is tedious and there is no reliable method at this stage. 2) Creating computational models which are controlled by a few parameters. One of the major problem is to find such models and to compose them to create complex motion. Such models can be created for walking, grasping, but also for groups and crowds.