AI of the same opinion: why neurocasters have become a trend
Every hour, four or five neurocasters appear in Runet - reworkings of popular songs made with the help of artificial intelligence. Experts note that the trend, which was born last year, has strengthened and will continue to develop in 2025, and many bloggers are already using this genre for commercial purposes. However, the trend is hampered by the complexity of creating such recordings and the lack of legal regulation, because of which large social networks and streaming services are not yet able to join the game and offer publicly available tools. "Izvestia" listened to the musical creativity of AI and sorted out its prospects.
From Cobain to Shufutinsky
In 2023, the song Heart on My Sleeve, "sung" by Drake and The Weeknd, became a sensation: only in TikTok it was listened to more than 15 million times. In fact, neither of the two artists had anything to do with it: their voices were generated with the help of a neural network. The composition was removed from YouTube due to claims of Universal Music Group, the holding company working with both artists. The apotheosis of the scandal was the statement of the general director of the American National Academy of Recording Arts and Sciences Harvey Mason, Jr. who allowed Heart on My Sleeve to be nominated for a Grammy. Technically, the rules of the award were not broken, he said, because the author of the music and lyrics is a real person (hiding under the pseudonym Ghostwriter977) and he did not steal his work from anyone. But very soon Mason denied himself and ruled out the possibility of nomination due to legal contradictions.
Anyway, the trend was created, and new compositions with the use of AI began to appear like mushrooms after the rain, including in Runet. And users began to master a variety of approaches. And if Heart on My Sleeve really was original material, then many later examples used someone else's previously created music or poems (and more often both). For example, more than 500 thousand views were gathered by a recording where Kurt Cobain's voice sings the lyrics of SHAMAN's song "I'm Russian" to the melody of Nirvana's famous hit Smells Like Teen Spirit. Anna Asti's "Tsarina," allegedly performed by Metallica, was even more viral: it was listened to more than 3.6 million times on YouTube alone.
On various services today there are channels dedicated exclusively to neurocasters, often focused on one style or artist. For example, the AI COVEROCK Rutube channel features AI-created hard rock versions of the most unexpected songs: the brutal power of metal has already been found in "Antoshka" and "Blue Tractor", "Chastushki babo-ezhka" from "Flying Ship" and even Mikhail Shufutinsky's "Third of September". Naturally, all this is even more vulnerable in legal terms than one's own music "sung" by other people's voices. It is still unclear how to solve the problem.
Outside the law
Natalia Polianchik, acopyright expert, CEO of IPCodex law firm, associate arbitrator of the Arbitration Center at the Russian Union of Industrialists and Entrepreneurs (RSPP),admits that the legislation does not keep up with theachievements of science in the development of neural networks. In a conversation with Izvestia, she noted that the creation of covers requires not only the authors' permission for a "new performance" of a song, but also the permission of the performers or their heirs for the results of the neural network's work to be used for public information.
- Creators and developers of artificial intelligence programs "feed" their children huge amounts of information, including phonograms with the voices of famous performers, under the pretext of educational purposes. But, as we see, products created by neural networks are increasingly being used to attract attention, advertising and promotion," the expert points out.
Even if the material used is a composition that has passed into the public domain, and it is "sung" in the voice of an artist who is no longer alive - for example, Joseph Kobzon "performs" "Murka" - there may be a claim from his heirs.
- Russian law guarantees protection of personal dignity, personal inviolability, honor and good name, business reputation, privacy, and the name of a citizen. Of course, if Kobzon's heirs go to court and prove that the use of the great singer's voice in a neurocaster violates the rights guaranteed by law, they will be able to prohibit the use of such "performances," summarized Natalia Polianchik.
However, living Russian performers seem to have taken a wait-and-see attitude towards the phenomenon. They do not take any legal steps (at least in the public sphere), but also do not support the folk reworking of their creativity or the use of their own voice. The creators of the neurocover of "Tsarina" reported that they offered Anna Asti to publish the track on streaming services, but to no avail. SHAMAN also knows about the popularity of "I am Russian" in the "performance" of Cobain, "Izvestia" found out. However, he refused to express his opinion about it and assess the trend in general.
Cover every 15 minutes
On the one hand, it is clear that popular covers once again draw attention to the artists themselves. On the other hand, because of the legal lack of regulation, it is problematic to make money on it (at least for a popular person, not an anonymous blogger), and the artistic level of such crafts is extremely uneven. Sometimes it can be done creatively and wittily, but more often the level is much lower. Which is not surprising, considering the number of tracks appearing.
- According to our data, only in the Russian segment of the Internet, four or five videos with AI covers of different quality appear every hour," Maxim Petrenchuk, head of the Center for the Study of Opinion Leaders and New Media, told Izvestia.
According to the expert, the trend will continue in 2025, but its growth will depend on whether the covers will be able to surprise the public. And, of course, it is necessary to regulate the sphere in terms of copyright.
The latter is also a stumbling block in that the major players in the Internet business will now have to pretend that the genre does not exist. Consequently, the production of new compositions will remain the domain of enthusiasts who are ready to understand the instructions and master specialized online services.
As a rule, Google Colab is used, as well as one of the sites that can separate vocals from instrumental (by the way, also with the help of AI), and a set of ready-made voices of artists. This is enough to "replace" the original vocal with another one, having only a regular MP3 in hand - without minus tracks and the like. But even such a result requires at least a fairly powerful computer and at least half an hour of time. More creative operations, such as changing the style of the song, melody, lyrics, and so on, and so on, bring the task to another level of complexity.
- The first social network to introduce the ability to create AI music will have a serious technological advantage for audience growth. This is how TikTok shot up in its time, because it provided ready-made templates that could be inserted into videos without having to work with editing programs," says Maxim Petrenchuk.
This can be compared to creating stickers for messengers. A few years ago, in order to get your own sticker, you had to have at least computer photo editors, but today any schoolchild can do the same, all you need is a smartphone and one minute of time.
Tool and intelligence
However, if stickers exist primarily in the space of personal correspondence, neurocasters are oriented to an external audience, and in addition to copyright issues, a whole range of moral and ethical issues arise here. In this regard, a comparison with dipfakes - using the appearance of people without their consent to create video content - is more appropriate. There are also fears that AI will simply supplant professional musicians in the foreseeable future. Why wait for Deep Purple to release an album of remakes from the 1940s-1960s (they have such experience: Turning To Crime), when you can do something similar yourself, and choose which songs you want to hear "performed" by Ian Gillan?
However, experts urge not to exaggerate the danger.
- The neural network revolution in music partly resembles the revolution of electric musical instruments, sound recording and sound amplification of the 20th century: there is millions of times more content, new institutions and styles have appeared. But, as then, fears about replacing the author and performer with a soulless iron are groundless, - Evgeny Safronov, editor-in-chief of the InterMedia news agency, is sure. - The essence of art is communication between people with the help of artistic images created by authors. The tool helps to create and transmit images, but their creation is the embodiment of the will and desire of the creator, and tools do not possess this.
According to him, neural networks will be applied in the industry more and more widely, actively helping authors, but in no way replacing them.
It is hard not to agree, at least if we are talking about real art, which is always based on an extraordinary view of the world: even the smartest AI is not capable of this, perhaps not yet. But can the very idea of combining the unconnected by letting Metallica sing "Antoshka" and Kobel's "Antoshka" be considered such an extraordinary view and a manifestation of talent? "Antoshka" and Cobain's "I'm Russian," is an open question.
In any case, having found a strange composition with the voice of your favorite singer on the web, now it's better to be wary and do not rush to share the novelty in social networks. After all, it will be a shame if artificial intelligence leaves the bearer of natural intelligence a fool.
Переведено сервисом «Яндекс Переводчик»
