In this paper we will be describing a new animation architecture and its implementation in our system LIVE. This model introduces a new blending layer approach which uses several motion generators on the same character at the same time. Despite the numerous studies done, animating characters or complex objects in a virtual world remains a difficult problem. In fact, the complexity of the data used to represent the characters (often articulated rigid bodies) makes their control by an animator difficult and using only one technique to generate motion tends to limit the quality of the animation. The solution will probably be provided by the co-operative or concurrency use of several motion control methods.