In the present paper, we introduce a new control architecture aimed at driving virtual humans in interaction with virtual environments, by motion capture. It brings decoupling of functionalities, and also of stability thanks to passivity.We show projections can break passivity, and thus must be used carefully.Our control scheme enables task space and internal control, contact, and joint limits management. Thanks to passivity, it can be easily extended.Besides, we introduce a new tool as for manikin's control, which makes it able to build passive projections, so as to guide the virtual manikin when sharp movements are needed.