A highly interpretable, controllable, and flexible framework for hierarchical melody generation
My group is collaborating with the Duke Office for Translation and Commercialization to license our music generation models.
A novel representation for musical sentiment and an interpretable model for melody harmonization. Used in conjunction with SchenkComposer for our NeurIPS creative track presentation.
The largest dataset of Schenkerian analyses in a computer-readable format.
Email stephen (dot) hahn (at) duke (dot) edu for access to over 140 analyses.
The first notation software for Schenkerian analysis, easily translating between JSON and human-readable formats. Analyze directly on the page with slurs and stems rendering in real-time.
Tools for music analysts and downstream music generation tasks that perform hierarchical music analysis based on our Schenkerian dataset.