Enacting and capturing real motion for all potential scenarios is prohibitively expensive; hence, there is a great demand to synthetically generate realistic human motion. However, it is a central challenge in character animation to synthetically generate a large sequence of smooth human motion. We present a novel, database-centric solution to address this challenge. We demonstrate a method of generating long sequences of motion by performing various similarity-based "joins" on a database of captured motion sequences. This demo illustrates our system (MoDB) and showcases the process of encoding captured motion into relational data and generating realistic motion by concatenating sub-sequences of the captured data according to feasibility metrics. The demo features an interactive character that moves towards user-specified targets; the character's motion is generated by relying on the real time performance of the database for indexing and selection of feasible sub-sequences.