Why Producers IllMind & Cam O’bi Are Experimenting With AI in the Studio

With a handful of exceptions, the era of rap’s six-figure super producer – when Scott Storch, Just Blaze, and Timbaland instrumentals effectively guaranteed radio hits – is over. The internet has made beat making more accessible than ever, and that diffusion means a Dutch teenager could craft one of the biggest records in music history for an unknown internet-savvy rapper dabbling in country tropes, or that a Romanian musician could become one of the go-to producers for Atlanta ragers Playboi Carti and Ken Carson. 

But even these seismic developments pale in comparison to what we’re seeing now: the beginning of generative AI in production and songwriting.

Given its technical nature, production has always been a component of music-making ripe for new developments, from the introduction of DAWs (Digital Audio Workstations) in the late 1970s by Soundstream to the advent of the MIDI digital composition format all the way to the rise of downloadable sample packs and eventually, the integration of AI. A July 2024 survey conducted by Tracklib found that 25 percent of producers were using AI in some capacity. The majority of those AI adopters were using it for either splitting stems or mastering, but more than 20 percent of them were also incorporating it into their records.

That 25 percent isn’t just beginners – it includes some of the genre’s most successful producers. In October, Timbaland was announced as a strategic advisor for Suno AI, a generative musical tool with some controversy around its use of copyrighted material to train its model. He even told Rolling Stone he spent 10 hours a day experimenting with the platform and writing text prompts to recreate some of his own legendary records. 

But the star producer who has most emphatically embraced AI is probably Illmind, a Grammy winner with credits for J. Cole, Beyoncé, and Jay-Z. Illmind has always been particularly ambitious when it comes to emerging technologies. He launched Blap Kits, his own sound sample platform, back in 2010, and started the AI-based LoopMagic earlier in 2024. With LoopMagic, musicians enter prompts into the interface for what sounds they want, and those are downloadable. (Subscriptions range from $25 a month up to $250.) They then own those sounds fully, per Illmind, though those in the lower-membership tier cannot package generated sounds as part of their own purchasable sample packs.

Illmind says that he was first introduced to AI as a musical tool back in 2016, when he experimented with a VST (virtual studio technology) plug-in that created MIDI chords and melodies. “I realized how powerful that was, because it was generating ideas from scratch that I musically wouldn’t normally think of,” he says. 

While AI’s depiction in the media could lead some to believe it is entirely autonomous, these programs need to be trained on something, and that’s one of the biggest ethical concerns in its widespread musical implementation. In September, a bill in California passed necessitating transparency with generative AI, and the Federal Trade Commission has targeted “unfair or deceptive practices” in the medium through its Operation AI Comply. Suno and another company called Udio have been sued by the three major labels, alleging “mass infringement of copyrighted sound recordings,” igniting a debate around whether AI model training should fall under the category of “fair use.”

Illmind’s LoopMagic was “only trained with permissioned data,” per its website, which includes Illmind’s sprawling catalog. The company Musical AI lists its manifesto as “Ethical Attribution in Generative AI Music Models,” and founders Sean Power and Matt Adell say that they’re taking their extensive history in music and tech and applying it towards creating a fairer world of generative AI. (Adell, the company’s COO, was an executive at Napster, so he’s certainly familiar with disruptive technology in the musical world.)

“Very quickly things balloon into needing a tremendous amount of data and we care for that a lot – because that data to be able to train on that vastness of information requires inputs, requires objects that you can’t possibly do without including copyrighted content,” Power says. 

Musical AI is trying to bridge the gap between existing musical creators and rights holders and AI companies, creating an ecosystem where artists are aware of any time their music is used to train AI and be financially compensated in those instances. Having worked through the period of illegal online downloads moving to paid downloads and eventually streaming, Adell says he thinks that the powers that be are comparatively responding much more quickly to the rise of AI, recognizing its urgency and transformational power for the music world.

“I actually think that industry agreements and government regulation are moving quite quickly compared to as they have in the past,” says Musical AI’s Adell. “When Sean and I started, we thought it could be five[-plus] years before all these mechanics get worked out – it [seems to] us now they’re going to get worked out in the next 18 months.”

Opinions on using generative AI to assist in music-making are decidedly split. Many defenders of the technology stress their belief that there will always be some human involvement in production and songwriting. Some artists, like Bay Area producer-vocalist Warren Long (a.k.a. Larrenwong) have embraced it, using programs like ChatGPT to help with lyrical ideas as well as the occasional musical program to aid in production, though he says the latter is often underwhelming and hasn’t frequently made it into his finished work. Still, he talks about AI in music creation with an emphatic pragmatism.

“I’m the first person to tell somebody to use AI for anything,” Long says. “I’m a proponent of it. We always read back and hear the case studies of people, they don’t take advantage of [emerging technology] and then they get left behind.”

Long says he has minimal “ethical qualms”  around the use of AI in his own music. “At this point? If it sounds good, f–k it,” Long explains – though he says he’d be unlikely to use an AI program that took a percentage of his rights and royalties. So much of tech innovation has been driven by financial shrewdness, and in this brutal creative economy, the allure of a cheap tool for production or songwriting is similar to the appeal many industries are navigating with AI-led automation.

All of this connects to a kind of existential question in production and songwriting: how much conventional musical work should a music-maker be doing? Rap producers have always faced criticism on this front, from those who derided sampling, to critiques of using premade loops, to now this discussion around AI. A rap producer who advocates for AI could easily make the argument that critical resistance to generative AI is the modern equivalent of traditionalist listeners balking at Grandmaster Flash or the Beastie Boys building now-iconic records through sampling.

The difference here is that human artists were flipping those samples into new songs themselves, whereas these AI platforms are algorithmically combing through musical data with the only real input typically being a written prompt. Cam O’bi, best known as a producer for Chance the Rapper and Noname, addresses the question of whether software that’s trained on existing music without the ephemeral secret ingredient of human creativity can ever create anything truly new or innovative. His skepticism comes in part from a belief that the generative AI software can’t truly innovate, and will instead come up with simplistic regurgitations of the music that it intakes, not unlike how a green musician often struggles to create novel sounds and styles from what they’ve learned. 

“When a person is learning how to make music or play guitar and they just keep f–king sounding like Jimi Hendrix, that just means that they still have learning to do,” says O’bi, who also has a budding solo career. “They have to get better, so that they can figure out how to incorporate Jimi Hendrix into their playing without just becoming a cheap imitation of him.”

O’bi explains that he has tried out different AI tools in his work, specifically for stem separation and to craft character voices that would appear on skits throughout one of his albums. (Attempting to combine his own voice with a pirate voice from the video game League of Legends, he found the results “unusable” due to poor audio quality, and the inability to direct the line reading. O’bi says he’s experimented with using ChatGPT like a rhyming dictionary to help move ideas forward, but has been consistently underwhelmed.

“ChatGPT is not a great writer,” O’bi says simply.

Illmind has used his LoopMagic AI program in sessions with other producers, including a widely circulated clip of himself and producer Dunk Rock (Gunna’s “fukumean” and “Cooler Than a Bitch”) generating a “Mysterious sample in F minor.” Even in that video, he does clarify that the software “was ethically trained by me,” and seems a little sheepish about the whole process, though perhaps just for comedic effect. Where it counts though, he’s all-in on the generative AI musical revolution, saying his goal is for LoopMagic to be the “ultimate co-producer.”

“There’s this quote that I read recently and it said, ‘AI won’t replace your job. The human beings that use AI will,’” Illmind says.

It’s not hard to picture a world where stem splitting and other rote, tedious tasks a producer must do for themselves are outsourced almost entirely to AI. Stem splitting, the act of separating the individual audio tracks of a song, typically to isolate vocals or a single instrument part, doesn’t fall under generative AI, but the more conventional use of the technology. It’s an important part of hip-hop production, but it has a different level of sacredness than the process of taking your blank screen and turning it into a track. Being able to use a generative AI tool is absolutely a skill, but it doesn’t seem like one that will make its way into every producer’s toolbox.

Obi says that he’d rather hire another co-producer to help with the workload – like a Dr. Dre or Kanye West has in the past – than to outsource key elements of his production work to AI technology: “Creating the music is really the most rewarding part. And the most fun part.”

Related Images: