Artificial intelligence composers will soon be scoring videos with a new tool from photo- and video-editing platform PicsArt that taps the latest research on machine-learning-powered music generation.
The feature, rolled out in partnership with AI music startup Mubert, will generate an instrumental track in any of more than two dozen musical styles designed to match the mood of a given clip. Trained on a library of over 1 million beats, samples and patterns, the system also has the benefit of being completely royalty-free, since each snippet it generates is a wholly original composition.
“AI, of course, cannot replace human imagination, but it will become the new virtual creative assistant that will expand and fine-tune one’s overall skills,” PicsArt CEO Hovhannes Avoyan said. “In time, we’ll begin to see more novice users being able to produce vastly superior content than they ever thought possible because of AI.”
Among the genres of music offered by the feature are “techno,” “lofi” and “dub” as well as mood-based styles including”happy” or “romantic” and activity-based ones such as “study” or “yoga.” If the output is not to the user’s liking, they can simply hit “regenerate” and it will produce another track in the same style.
The new tool comes weeks after research group OpenAI released a system called Jukebox, which is able to generate realistic music and lyrics in the style of various hyper-specific pop genres. Agencies have also begun to jump on the AI music trend; agency Space150 created a passable imitation of a Travis Scott song and music video earlier this year, and Adidas has also contracted with Mubert on a campaign that featured AI-generated tracks.
Despite the appeal of circumventing royalty fees, legal challenges may still attempt to root out gray areas in this emerging field. Jay-Z’s record label recently sent a cease-and-desist letter to a YouTube channel that does deepfake impressions of the rapper and other celebrities, claiming it was using AI to unlawfully impersonate him.