Stroke patients with hemiparesis often exhibit an asymmetrical weight distribution, with the paretic leg providing less support and less weight-shift activity than the sound leg does. For upper-limb hemiparesis constraint-induced movement therapy (CIMT), where the sound limb is constrained and the patient is forced to use the paretic limb, has proven to be one of the most effective rehabilitation methods. However, this method has not been successfully modified for balance rehabilitation, since hard constraints on the lower limb are difficult to impose and can increase the risk of patient injuries. Here we introduce a novel approach that encourages stroke survivors to use the paretic leg for weight-shifting tasks during balance control training in a virtual reality setting. We recorded motion signals from each leg in real time and streamed modified versions of these signals to the Nintendo Wii Fit gaming system. By independently manipulating the influence of each leg during game play, we successfully forced the stroke survivors to increase the use of the paretic leg. Our preliminary clinical trial, including three patients with hemiparesis in the chronic phase, found that one-week of training improved patients’ ability to maneuver their center of pressure (COP) during a tracking task. More encouragingly, patient’s weight distributions became more symmetrical. These initial results suggest that manipulating control gains during VR rehabilitation may allow the principles of CIMT to be applied during balance training in stroke survivors.