SynthAssist: Querying an Audio Synthesizer by Vocal Imitation

Cartwright, Mark and Pardo, Bryan

Proceedings of the International Conference on New Interfaces for Musical Expression

Programming an audio synthesizer can be a difficult task for many. However, if a user has a general idea of the sound they are trying to program, they may be able to imitate it with their voice. This paper presents SynthAssist, a system for interactively searching the synthesis space of an audio synthesizer. In this work, we present how to use the system for querying a database of audio synthesizer patches (i.e. settings/parameters) by vocal imitation and user feedback. To account for the limitations of the human voice, it uses both absolute and relative time series representations of features and relevance feedback on both the feature weights and time series to refine the query. The method presented in this paper can be used to search through large databases of previously existing “factory presets” or program a synthesizer using the data-driven approach to automatic synthesizer programming.