Multi-Source Diffusion Models for Simultaneous Music Generation and Separation

arXiv:2302.02257v4 Announce Type: replace-cross
Abstract: In this work, we define a diffusion-based generative model capable of both music synthesis and source separation by learning the score of the joint probability density of sources sharing a context. Alongside the classic total inference tasks (i.e., generating a mixture, separating the sources), we also introduce and experiment on the partial generation task of source imputation, where we generate a subset of the sources given the others (e.g., play a piano track that goes …

abstract arxiv context cs.lg cs.sd diffusion diffusion models eess.as generative inference music music generation music synthesis probability synthesis tasks total type work