Reproduce “Multi-Source Diffusion Models for Simultaneous Music Generation and Separation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Giorgio Mariani, Irene Tallini, Emilian Postolache, Michele Mancusi, Luca Cosmo, Emanuele Rodolà
The authors train one diffusion model on coordinated instrument tracks, using it for full generation, partial generation, and source separation. On Slakh2100, the model scores FAD 6.55, close to 6.67 for a mixture-trained counterpart. For separation, corrected MSDM with Dirac likelihood scores 16.48 dB overall, below Demucs plus Gibbs sampling at 17.73 dB, on four instrument classes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.