Supporting Interactive Machine Learning Approaches to Building Musical Instruments in the Browser

Interactive machine learning (IML) is an approach to building interactive systems, including DMIs, focusing on iterative end-user data provision and direct evaluation. This paper describes the implementation of a Javascript library, encapsulating many of the boilerplate needs of building IML systems for creative tasks with minimal code inclusion and low barrier to entry. Further, we present a set of complimentary Audio Worklet-backed instruments to allow for in-browser creation of new musical systems able to run concurrently with various computationally expensive feature extractor and lightweight machine learning models without the interference often seen in interactive Web Audio applications.