MusicLM is an innovative AI model developed by Google Research that transforms detailed text descriptions into high-quality music.
By interpreting prompts like “a calming violin melody backed by a distorted guitar riff,” it produces audio at 24 kHz, ensuring clarity and consistency over extended durations. This capability allows users to create diverse musical pieces simply through textual input.
A standout feature of MusicLM is its ability to incorporate both text and melody conditioning. This means you can input a whistled or hummed tune, and the model will adapt it to match the style outlined in your text description.
To support further research and development, Google has released MusicCaps, a dataset containing 5,500 music-text pairs with detailed descriptions provided by experts.