In this article, we present a technical framework aimed at facilitating musical biofeedback research in poststroke movement rehabilitation. The framework comprises wireless wearable inertial sensors and software built with inexpensive and opensource tools. The software enables layered and adjustable music synthesis and has a generic movement-music mapping module.Using this, we designed digital musical interactions for balance, sit-to-stand, and gait training. Preliminary trials with subacute stroke patients indicated that the interactions were clinically feasible. Expert interviews with a focus group of clinicians were also conducted, where these interactions were deemed as meaningful and relevant to clinical protocols, with comprehensible feedback (albeit sometimes unpleasant or disturbing) for several patient types. We carried out system benchmarking, finding that the system has sufficiently short loop delays (∼90 ms) and a healthy sensing range (>9 m) and is computationally efficient (11.1% peak CPU usage on a quad-core processor). Future studies will focus on using this framework with patients to both develop the interactions further and measure their effects on motor learning, performance retention, and psychological factors to help gauge their true clinical potential.