music

Definition

Music as a modelling target, distinct from speech or general audio. What makes it hard is long-range structure: beat, key, and sectional form must stay coherent across tens of seconds to minutes, a horizon where sample-level generative models drift. It is also the modality where training-data rights most directly shape what gets released — several prominent systems have been described in papers while the weights and data stayed closed.

Explainers using this term

• Music and Audio Generation from Scratch — Sound as TokensSimple and Controllable Music Generation