No title
Music generation has progressed rapidly in recent years, but practical track-to-track transitions are still largely handled with conventional signal processing such as equal-power crossfades and manual beatmatching. This thesis evaluates whether latent-space interpolation in a pretrained neural audio codec can serve as a short, targeted generation operator for music transitions, where the goal is
