MusicGen-Chord: Advancing Music Generation through Chord Progressions and Interactive Web-UI
Jongmin Jung, Andreas Jansson, Dasaem Jeong
TL;DR
MusicGen-Chord tackles harmonic conditioning in neural music generation by representing chord progressions with multi-hot chroma vectors and conditioning a pretrained transformer-based MusicGen model without fine-tuning. The approach supports both text-based and audio-based chord inputs, enabling chord-aware generation that aligns with user prompts and harmonic structure. The companion MusicGen-Remixer workflow integrates BPM/downbeat analysis, Demucs source separation, BTC chord extraction, Dynamic Time Warping, and mixing to produce coherent remixes that fuse the input audio with a regenerated background. Deployment via Replicate's cog-based web-UI makes these capabilities broadly accessible, illustrating a scalable, user-friendly path for cloud-based AI-driven music creation and remixing.
Abstract
MusicGen is a music generation language model (LM) that can be conditioned on textual descriptions and melodic features. We introduce MusicGen-Chord, which extends this capability by incorporating chord progression features. This model modifies one-hot encoded melody chroma vectors into multi-hot encoded chord chroma vectors, enabling the generation of music that reflects both chord progressions and textual descriptions. Furthermore, we developed MusicGen-Remixer, an application utilizing MusicGen-Chord to generate remixes of input music conditioned on textual descriptions. Both models are integrated into Replicate's web-UI using cog, facilitating broad accessibility and user-friendly controllable interaction for creating and experiencing AI-generated music.