Check out MusicLM, the Artificial Intelligence web app that creates high-quality music from text prompts if you enjoy music. The app elevates the process of creating conditional music to a whole new level with its hierarchical sequence-to-sequence modeling. In terms of audio quality and adherence to the text description, the app's developers, Andrea Agostinelli, Timo I. Denk, Zalán Borsos, Jesse Engel, Mauro Verzetti, Qingqing Huang, Aren Jansen, Marco Tagliasacchi, Matt Sharifi, Neil Zeghidour, and Christian Frank, did a fantastic job producing music at 24 kHz that outperforms earlier systems.

The ability of MusicLM to turn whistled and hummed melodies into music that follows the style specified in a text caption is one of its most intriguing features. You can create music that complies with text prompts while following a provided melody by adding melody embeddings to conditioning with MusicLM.

The creators kindly made MusicCaps, a dataset made up of 5.5k music-text pairs with extensive text descriptions provided by human experts, publicly available to support further research.

Rich text captions can be used to create music using MusicLM.