As the field of neural prosthetics advances, Brain Machine Interface (BMI) design requires the development of virtual prostheses that allow decoding algorithms to be tested for efficacy in a time- and cost-efficient manner. Using an x-ray and MRI-guided skeletal reconstruction, and a graphic artist's rendering of an anatomically correct macaque upper limb, we created a virtual avatar capable of independent movement across 27 degrees-of-freedom (DOF). Using a custom software interface, we animated the avatar's movements in real-time using kinematic data acquired from awake, behaving macaque subjects using a 16 camera motion capture system. Using this system, we demonstrate real-time, closed-loop control of up to 27 DOFs in a virtual prosthetic device. Thus, we describe a practical method of testing the efficacy of high-complexity BMI decoding algorithms without the expense of fabricating a physical prosthetic.