Beyond Music Information Retrieval: A Proposed Model for Automatic Generation of Dialogic Music

Resumo

In this article we propose a generative music model that recombines heterogeneous corpora of audio units on both horizontal and vertical dimensions of musical structure. In detail, we describe a system that relies on algorithmic strategies from the feld of music information retrieval—in particular content-based audio processing strategies—to infer information from music recordings that, in turn, supervise generative music strategies. The model allows automatic remix and mashup creation and the manipulation of audio signals according to inferred stylistic features. In addition to widening the creative potential of non-experts, the model also expands knowledge in areas such as computational music analysis, generative music and music information retrieval.