Content creation for computer graphics applications is a laborious process that requires skilled personnel. One fundamental problem is that manipulation of 3D objects with 2D user interfaces is very difficult for non-experienced users. In this paper, we describe a system that uses semantic constraints to restrict object motion in a 3D scene, making interaction much simpler and more intuitive. We compare three different levels of semantic constraints in a 3D scene manipulation program with a 2D user interface. We show that the pr esented techniques are significantly more efficient than alternate techniques, which do not use semantics in their constraints. To our knowledge, this is the first evaluation of 3D manipulation techniques with 2D devices and constraints.