MusicLM: Generating Music From Text
This addresses the challenge of creating music from text for applications in entertainment and creative tools, representing a novel method rather than an incremental improvement.
MusicLM tackles the problem of generating high-fidelity music from text descriptions, achieving consistent audio quality at 24 kHz and outperforming previous systems in both audio quality and text adherence.
We introduce MusicLM, a model generating high-fidelity music from text descriptions such as "a calming violin melody backed by a distorted guitar riff". MusicLM casts the process of conditional music generation as a hierarchical sequence-to-sequence modeling task, and it generates music at 24 kHz that remains consistent over several minutes. Our experiments show that MusicLM outperforms previous systems both in audio quality and adherence to the text description. Moreover, we demonstrate that MusicLM can be conditioned on both text and a melody in that it can transform whistled and hummed melodies according to the style described in a text caption. To support future research, we publicly release MusicCaps, a dataset composed of 5.5k music-text pairs, with rich text descriptions provided by human experts.