As AI slowly makes its way into every part of our digital lives, from writing essays to talking to therapists to creating original art, it doesn’t come as much of a surprise that it would also make its way into music.
Google has made a new artificial intelligence (AI) system that can make any music from a written description. Users could type in increasingly specific cues, such as genres and styles, or even build songs based on a hummed or whistled melody. The app will be called MusicLM inside the company.
Even though MusicLM wouldn’t be the first generative AI system for music, it is the first to make songs with “high-fidelity” and complicated composition.
After training on a set of 280,000 hours of music, the algorithm can make songs that make sense for words like “substantial complexity.”
The system can add to melodies that have already been played, whistled, hummed, sung, or played on an instrument.
Researchers at Google say that it can also turn a set of sequentially written descriptions into a musical “story” or narrative.
A picture and a caption can also control MusicLM. It can also make music that sounds like it was “played” by a certain kind of instrument in a certain style.
Even though the system can technically synthesize vocals, the results are far from perfect and have issues, such as samples that need clarification.
Copyright issues
The main problem for Google is that MusicLM might use training data with copyrighted material in the songs it makes.