Simple and Controllable Music Generation
This work addresses the problem of controllable music generation for AI and creative applications, offering a simpler and more efficient approach compared to prior hierarchical methods.
The paper tackles conditional music generation by introducing MusicGen, a single-stage transformer language model that uses compressed discrete music representations and efficient token interleaving to eliminate the need for cascading models. It demonstrates high-quality mono and stereo music generation conditioned on text or melody, showing superiority over baselines in automatic and human evaluations on a standard benchmark.
We tackle the task of conditional music generation. We introduce MusicGen, a single Language Model (LM) that operates over several streams of compressed discrete music representation, i.e., tokens. Unlike prior work, MusicGen is comprised of a single-stage transformer LM together with efficient token interleaving patterns, which eliminates the need for cascading several models, e.g., hierarchically or upsampling. Following this approach, we demonstrate how MusicGen can generate high-quality samples, both mono and stereo, while being conditioned on textual description or melodic features, allowing better controls over the generated output. We conduct extensive empirical evaluation, considering both automatic and human studies, showing the proposed approach is superior to the evaluated baselines on a standard text-to-music benchmark. Through ablation studies, we shed light over the importance of each of the components comprising MusicGen. Music samples, code, and models are available at https://github.com/facebookresearch/audiocraft