Users can simply type in a text prompt, like “smooth jazz” and MusicLM will generate two songs to fit the parameters. You then vote on which track you like best, helping the AI tool to learn what is considered “better music.”
You can even hum a melody or give it inspiration based on places, eras, and physical artwork.
Music LM was trained using 5 million audio clips with captions written by musicians that describe the music’s genre, mood, and overall feel.
Google did preview MusicLM in an academic paper in January, but said that due to ethical issues, they were not going to release the program. But after working with musicians and seeing how AI technology can help creativity, they decided not to use specific artists or vocals in MusicLM to hopefully reduce any copywriting issues.
But rest assured, musicians may not be out of a job with programs like these yet. A new study from the University of York found that AI-generated music may be less favored and inferior to human-composed music.
To learn more about text-to-music programs, check out the video below and check out Google’s MusicLM, a new AI music-generating platform, here.
Inspired by: edmmaniac.com