Music producers and a lawyer specializing in intellectual property talk to Gama about the use of AI tools to recreate voices, invent duets and develop new lyrics and instruments
From the cover of "Escândalo Íntimo", singer Luísa Sonza's new album, to the creation of "Heart on My Sleeve", a hit produced with the (fake) voices of The Weeknd and Drake, passing through the late Frank Sinatra (1915-1998) singing "Toxic" - albeit in a somewhat slurred voice -, a song Britney Spears released in 2003, and the meme that hit TikTok with Ariana Grande singing the lyrics to Zé Felipe's rap, artificial intelligence is now in all areas of music, from the beginning to the end of the industry's processes.
The exaggerated use of the various AI tools that appear every day in the music market, however, raises some relevant questions. Are there professionals in this field who could lose their jobs to machines? How legal and reasonable is it to use AI applications to produce songs? What about copyright? Will the world of pop music become even more pasteurized?
With these and other questions in mind, Gama consulted music producers and a lawyer who specializes in intellectual property.
Uses and benefits of artificial intelligence in music
Compared to the current artificial intelligence technologies that exist for creating videos, which are more developed and popularized, AI for music, especially for the creative side, is still moving more slowly.
A DJ, composer and music producer, Mulú says that technically, AI algorithms are already widely used in Brazil to mix and master or to separate all the instruments from the voice in a recording, for example. However, "creatively, it's something relatively new, with lots of possibilities to be explored, such as simulating the voice of artists who have already died," says Mulú, who has produced works by the likes of Duda Beat, Letrux, Luedji Luna and Pabllo Vittar.
I believe it's going to be an industry standard, mainly because there's a huge cost reduction
Another way of inserting AI into producers' daily lives is in pre-production, when these professionals are developing so-called demo tapes - demonstration recordings for studies or proposals of what could become a song or even an album. "Sometimes you're making a demo and you think that a certain song might be good for a certain singer, and using these algorithms that simulate voices, you can get a good idea beforehand," he says.
This prior virtual test, according to Mulú, would make the process of going to the studio easier, by advancing a stage. "I think this is going to be an industry standard in a few years' time, mainly because there's a huge reduction in the cost of taking the artist to the studio to record the voice, to find out if it's going to be OK. This way, you can make a pre-product and then record it for real."
Music producer João Marcello Bôscoli says that of the ten pieces of software he buys for his studio, seven must have artificial intelligence. In other words, these systems are here to stay. "It's a method, a way, a technology, and it's good to remember that artificial intelligence is just a generic term for thousands of things."
Bôscoli mentions that one of the uses he makes of AI is in the restoration of old recordings, with a program that removes any unwanted background noise. "The software identifies all the harmonics, all the manifestations of that particular group of noises, and you decide whether you want them to appear there or not. If you don't, just press a button and that background is suppressed. The voice is dry, as if it had been recorded in an acoustically sound room. And all this at a speed and with a precision that would not be possible for a human being to achieve," he says.
Use, but in moderation!
This function reported by Bôscoli is carried out by AI, but is not entirely completed by machine intelligence. In order to work, the application needs human commands. And it is this point in particular that the music producer is cautious about. "You have to be careful with the degree of interference from artificial intelligence, there's a minimum and a maximum," he says.
The highest intensity, he says, would be the creation of a fake artist, someone who can't sing but who, using listed references, "becomes" a singer.
"You come into the studio, say you like Lady Gaga and Olivia Rodrigo, I take that information, put it into the computer, press a button and I get a ready-made base in seconds. Then I say I want a voice 'like this', a melody 'baked', and that's it," he analyzes. "The artificial intelligence that helps with a restoration has one level, the AI that fixes the voice has another level and the one that generates music and lyrics is on another level."
João Marcello Bôscoli refers to a speech made by musician Gilberto Gil on the program "Roda Viva" on May 23, 2022, when he was asked what the maximum line would be for the use of technology in musical creation. And the veteran's answer was: "I think that for the arts in general, the limit is the human command".
I don't see the gain when artificial intelligence is taking my creative work process away from me
Bôscoli's thinking follows the same line. He uses AIs as facilitators of production phases, and doesn't see any problems with this, but it doesn't cross the line into creativity. "I don't see the gain when artificial intelligence is taking my creative work process away from me. The thing I enjoy most in life is making music. I use artificial intelligence in various stages of production, no problem, but everything that is part of the creative process I don't give up, just as I don't give up cooking for my children," he explains.
On a day-to-day basis, he says, it's not always possible to go into the kitchen, so he resorts to lunchboxes that he buys and leaves in the freezer. "It's just not the same, I'm feeding them both, the food is good, organic, and everything is fine, but I prefer to choose, cut and make it myself, for the pleasure of it. So I use artificial intelligence for all the things I couldn't or wouldn't be able to do, as a creative tool, but I don't want it to replace me."
The music producer also touches on the question of the future of work, since AI is fed with data that we pass on with each use. "So it's about using it as a model, as a tool for your own creativity. The important thing is not to let yourself be used by artificial intelligence. It's a cliché, grandpa's advice, but every time you use AI, every time you talk to it, you're giving the tool information. And who's to say that in the future the contractor won't prefer the AI producer?" he asks.
Artificial intelligence, from lyrics to promotion
Owner of the hit "Michael Douglas", which exploded in 2016, João Brasil is an artificial intelligence enthusiast. When he made the song that repeated "Never again will I sleep, never again will I sleep. What's that? Michael Douglas!", he didn't use AI. Now, he has just released the funk "Take Me to the Stars", composed from start to finish by artificial intelligence platforms.
For the lyrics, Brasil accessed ChatGPT using commands like: "I want you to write a rap in English, with four lines, on the theme of space travel". Then, to set it to music, he passed the lyrics that ChatGPT came up with to the Splash Pro generator, which returned the song with a melody. To extract the "tamborzão" that appears in the funk, he used a program called Moises and then edited everything in Ableton Live.
The mastering was done at Ozone, while the cover artwork for the track was produced at Midjourney. Finally, the producer uploaded the content to SoundCloud Impossible Records AI.
João Brasil says that the machines are his partners and even though he uses their intelligence, the result doesn't come straight out of the box and, in addition, his references, just like those stored by the AIs, are also taken into account in this process.
"I like experimentalism. And it's obvious that everything I do comes from tests, mistakes and successes. If the lyrics of ChatGPT aren't to my taste, if something comes in that isn't related, I go and change a word, I adjust. You also use human intelligence. I think AI is a 'partner', I don't see any disadvantages," he explains.
Pasteurized pop scene?
"In a way, pop is already a bit pasteurized," says Mulú. But the producer believes that AI will help improve what already exists through a thousand and one possibilities, from the aforementioned, more popular technical uses to aesthetic creations, such as the appearance of unusual voices.
"The voice in music, in a song, is very important. I've seen the success of Marina Sena, who has a very peculiar voice, and Poze do Rodo too. That's why I believe that artificial intelligence will create very original voices. Not necessarily just copy voices that already exist," he says.
Artificial intelligence only enhances everyone's creative power, because you use technology to your advantage, dictating what you want
AI models, Mulú adds, can be trained to combine aspects of different voices. "That's one of the things that inspires me a lot, especially for instrument timbre, you can mix characteristics of different instruments, creating an instrument that doesn't even exist acoustically."
João Brasil agrees. According to him, AIs increase creativity because technology is used in favor of those who produce. "If you want to make current pop, you go to the prompt [a message that an operating system sends to ask you to enter information to execute a command] and write that you want it. An experimental artist says what he's looking for and the tool goes to an experimental side, with ideas for melodies to be played on the saxophone, for example. So I think artificial intelligence only enhances everyone's creative power, because you use technology to your advantage, dictating what you want. The machine still doesn't work completely on its own," he says.
In João Marcello Bôscoli's opinion, there is a tendency towards pasteurization of the scene, but in the end he thinks that human creativity will prevail over the machines, "with the discovery of different uses for this language".
He also points out that there is no style without reiteration. "There is no style without repetition. You recognize Miles Davis or Michael Jackson because they repeat patterns. But when the repetition is excessive, it starts to become 'ton sur ton', until it disappears into the landscape. But I think that both pop and funk people, the ones I know at least, are quite creative and I think they'll use these tools in a more inventive way."
I don't want them to like a computer that impersonates me
Again, his concern is not with a pasteurized system, but with the possibility of some professionals falling out of favor and no longer being hired because contractors, in a while, may prefer to buy software that will do the same job for a lower price and more quickly. After all, these programs have been fertilized (and taught) on a daily basis with content passed on by these experts.
"It's for this reason, because I mainly like this part of my work, that I don't use artificial intelligence in my musical creation. I'm not going to teach any bum robot to do what I know how to do, which may be bad or average, but it's what I know how to do, and some people will listen and maybe like it. So I'd rather they liked me. I don't want them to like a computer that impersonates me," he concludes.
Legal, moral and philosophical issues
Every day, technology advances and laws are not always able to keep up with these advances immediately. According to Luiz Fernando Plastino Andrade, a lawyer specializing in intellectual property, digital technologies and data protection, there is no specific legislation on artificial intelligence in Brazil.
There are, however, some bills being debated. "Initially, most of them focused on responsibility for artificial intelligence decisions. Now, with the popularization of the use of generative artificial intelligence, discussions have also been guided by issues of image rights, copyright, remuneration and the possibility of replacing some types of intellectual work with artificial intelligence," he lists.
Although there is still no specific legislation, Andrade explains that some existing laws regulate certain aspects that have arisen as a result of the use of AIs and that could affect society in some areas.
To this end, there is, for example, "the Copyright Law, the General Personal Data Protection Law, the Civil Code, which deals with image rights. We also have some provisions in the Federal Constitution that deal with aspects of personality. So all of this does apply to generative artificial intelligence as protection."
There are countless possibilities for using AIs, going as far as recreating the image and voice of people, in this case singers, who have already died.
The question is usually a moral one and revolves around wondering what that artist's will was. If there is no will on the subject, authorizing or not, even if the use is consented to by the family - or was approved in writing before the death - how do we know that singing in a particular duet, appearing on a concert screen or in a TV commercial would or would not be the will of the departed? For Mulú, these are philosophical discussions that humanity has not yet managed to reach."
Source: Gama Magazine