Streaming of kinect data for interactive in-browser audio performances

Distributed audio applications typically follow a server-client model, in which the server produces an audio stream that is sent to the client for consumption. Then, if any modification of the content is needed, it is a product of audio post-processing on the client side, which is computationally expensive. We tackle these challenges by separating the creative (input), from the sound processing (output) part. As a result, the application becomes lightweight in networking and computational metrics, while an extra level of control over the final output is added on the client. In order to demonstrate our solution, we present the architecture of a sample audio application that supports visualizations and uses Kinect as a controller.