Manual assembly of virtual 3D objects is required in several application fields. We focus on tangible user interfaces which offer the opportunity to the user to perform virtual assemblies efficiently and easily. In each hand, the user manipulates a tracked prop, and the translations and rotations are directly mapped to the corresponding virtual object. However, with such interfaces, both hands are requisitioned, and the user cannot drop the props without changing the action or the expected results. We list and discuss the choice of four possible modalities to activate/deactivate the assembly modality: vocal modality, gestural modality, buttons, and foot pedals. We conclude that when using the foot pedals, the user's gesture is closer to the real-world behaviour.