The Beatles resurrected with new tracks generated by artificial intelligence

It is a new version of New, a single by Paul McCartney released in 2013, on which a fan rejuvenated the voice of the Beatles and added a part sung by his former partner John Lennon, who died more than forty years ago. Equally impressive: a version of Grow Old With Me, one of the last songs written by John Lennon, released after his assassination in 1980 and recreated by an artificial intelligence enthusiast calling himself Dae Lims. The song’s most emotional moment comes when Paul McCartney starts singing along to the melody.

These creations moved many fans to tears. “I am in tears! It’s so beautiful!!!” wrote a listener on YouTube, under the first version. “I start to cry“, said in a video Steve Onotera, a youtuber with a million subscribers about the second. “When we see them brought together artificially but convincingly thanks to artificial intelligence, it is very moving“.

How are these AI versions made?

The technology used analyzes and captures the nuances of a voice. The creators probably sing the lyrics themselves, before applying a “clone” of the desired voice, like applying a filter to a photograph. Getting there is not easy and requires humans who know how to handle AI, with a real knowledge of music software, according to Zohaib Ahmed, CEO of Resemble AI, a Toronto company specializing in this field.

“I think there is still only a very small part of the population that can access these tools“, he said. We must”read on it, have the right computer, and put it all together“His company is one of those offering a platform that can be distributed more widely to the entertainment industry – including a so-called Netflix documentary”commented” by the voice of the late Andy Warhol.

For Patricia Alessandrini, composer and teacher at Stanford, the growing number of songs generated by AI represents the coming of age of a technology that has developed exponentially over the last decade, while remaining quite far from the great. audience. “This is a good example of what AI does really well, which is anything to do with resemblance“, she said. But”there is really no risk”, according to her“that it supersedes the rich history of art and culture created by humans.“

What about copyrights?

If these creations highlight recent technological advances, they also raise a host of ethical and legal questions. For the music industry, the implications are enormous. Technology allowing people to easily transform their voice into that of their favorite singer probably won’t be long in coming.

If the artistsare paid to license their voice, so everyone is happy“, believes Steve Onotera. “But what about those who are long dead?“

In the case of Heart on a Sleeve, a song that recently artificially brought together artists Drake and The Weeknd, the Universal group had the piece removed from streaming platforms – without preventing it from reappearing online here and there. According to Marc Ostrow, a lawyer specializing in these issues, AI-generated music is a “gray area“.

The industry will have to set standards

Rights can be claimed by both artists and music labels. But content creators using AI can claim the legal concept of “fair use” (“reasonable use“), a sort of exception clause. The American Supreme Court leaned in the opposite direction last month, ruling that a photographer, whose photo of the musician Prince was used by the painter Andy Warhol, should have received rights author.

“I think the standards will be deliberately set by the industry (…) or it will end up in litigation“, believes lawyer Marc Ostrow. Labels will indeed have to deal with the bad publicity caused by a lawsuit against the work of a fan, seen as a tribute and not a source of money.