If you’re done with the hype of ChatGPT, here’s something that may similarly excite you – an AI-based music generator from Google called MusicLM.
Research published by Google recently states that MusicLM can create songs of any genre with a text description. Based on the analysis of hundreds of thousands of music hours, this technology can even write abstract concepts that other regular AI generators can’t.
Google’s Effort to Tackle ChatGPT
As all around the world are mesmerized by ChatGPT’s work in various fields, Google came up with its AI model that can create surprising songs out of text. While it’s not the first-ever AI-based music generator, Google’s MusicLM is truly a different technology – based on its advanced results.
The research published on Google’s GitHub revealed that MusicLM could produce music with surprising variety and depth since it’s trained on a vast database of 280,000 hours of music. Well, aside from simply combining genres and instruments, MusicLM can even write tracks using abstract concepts that are usually difficult for general computers to produce.
Also, Google says that it can craft melodies based on humming, whistling, or just the text description of a painting! A new option called Story Mode can even stitch several descriptions together to produce a DJ set or soundtrack!
While it seems an advanced technology, it has its own share of problems too. For example, some output sounded strange, and vocals tended to be incomprehensible. Also, some can be repetitive in ways human works might not do – thus, still not better than humans.
And since it’s evident that AI generators need a bunch of copyrights to remake, they’re obviously violating the copyrights if released without appropriate permissions. Industry experts say that you may have to get clearance to release an AI-created song, much like musicians today, who base their final output based on pre-made samples.
