In one embodiment, a method includes identifying a three-dimensional gesture made by a user of a computing device with respect to one or more surfaces of the computing device, the three-dimensional gesture comprising a trajectory in three dimensions, a first portion of the trajectory comprising a touch of one or more of the surfaces, a second portion of the trajectory comprising a series of points in space distant from the surfaces; determining a user input based on the three-dimensional gesture; and executing one or more actions based on the user input.