What Does Sampling Mean in An Era of Instant Music Generation?

On the marketing materials for AI music generation platform Suno, the pitch is a service that’s clean and frictionless: “Everything you need to make music your way.” The promise from the Cambridge, Massachusetts-based company is immediacy and ownership. “Make up to 500 custom songs a month, with full commercial rights on the Pro plan,” the materials continue. The company describes itself in language that positions music generation as essential to self-expression. “Suno is a music company built to amplify imagination. Where ideas flow into sound and songs become yours.” Music-making through the platform is framed as a kind of instant translation from thought to product, ready for the market.

But nothing is ever really that easy or uncomplicated. Suno might aim to reshape artists’ workflow from plodding away at their instruments or sampler for hours on end to typing in a prompt and getting an entire finished piece of music, but the work and labour involved hasn’t been made easier, it’s just been deferred and made invisible. In this new model, users type a few words aiming at a specific vibe, mood, or style and receive a complete track, vocals included, with a license attached.

Sampling, one of the dominant modes of music production today, sits awkwardly beside this fantasy of effortlessness. The right sample can help a song come together quickly, but the process is rarely frictionless. It is an art of attention, of patience, of maintaining a closeness to musical mediums, be they tape, vinyl, or lines of code. It is a practice that is grounded in listening, searching, cutting, and arranging. But what does all of that mean or stand for when music no longer appears to require labour? Why bother searching for the perfect sample when you can just prompt one into existence? What do we lose when sound creation becomes instant, and when the story of music shifts from craft to convenience?

Sampling is building

Listen closely to how samplers talk about their craft and you start hearing words that evoke building and construction. Travis John, a young Toronto-based artist who has produced experimental hip hop as bridge of sand. for the past three years, describes his work in terms of mutable structures and built forms that emerge from his sampling choices. “Sampling was innately a part of why I wanted to make music,” he explains, citing acts like JPEGMAFIA, Earl Sweatshirt, and J Dilla as producers whose sample-based music had a tremendous impact on not just his desire to make his own music, but formed his understanding of how he could approach songwriting and production. “Hip hop is kind of the base route of how I learned to make music,” he says. “[Now,] I can take the structure that I've set up for myself with hip hop and apply it to different textural forms.” 

Edmonton-based musician Parker Thiessen echoes John’s sentiments. Thiessen describes his early approach in blunt, matter-of-fact terms: “[I was] just being like, ‘oh, this drum fits with this one, I’ll put those two together.’ That kind of thing. Like putting a puzzle piece together.” Like John, Thiessen’s musical origins also begin with hip hop, starting out as a turntablist in a rap group before branching off with his own productions in FL Studio. Getting turned on to artists like Kid Koala, Christian Marclay and Maria Chavez then opened him up to the experimental possibilities of turntablism. He has since founded numerous electronic projects that oftentimes centre on or involve sampling in some way. 

In music, “sampling” has come to name a technical process that encompasses multiple distinct cultural practices at once. The word was coined by Australian inventors Kim Ryrie and Peter Vogel, creators of the Fairlight CMI synthesizer, which included one of the earliest mass-market samplers available to consumers in its impressive array of functions. Introduced in 1979, the Fairlight made recording and replaying sound at different pitches feel like a new instrument, even with the limits of early digital memory and fidelity.

But the practice of transforming recorded sound into music did not begin there. Long before samplers sat in studios, musique concrète composers were recording everyday sounds onto tape, cutting and splicing them, manipulating their speed and direction, and using those “sound objects” to build compositions. There was Halim El-Dabh, who recorded the ambient sounds of Cairo’s streets with a wire recorder. A few years later, Pierre Schaeffer applied those same ideas to magnetic tape, and developed the theoretical basis of musique concrète in the early 1940s. 

Concurrently with—and independently from—what was developing in Cairo and Paris, Jamaican sound system operators treated record pressings as modular parts, selecting musical passages that could carry a crowd, then recontextualizing them through performance, repetition, and competition. Through a mix of happy accidents and innovation, that practice hardened into “versions,” dubplates, and later dub, where the record was treated as a multitrack object even when the source began as a finished song. This particular lineage paved the way for hip hop DJs and producers to turn breaks, fragments, and repeated bars into dance floor dominance and a form of storytelling that was distinct from academic and musicological currents. 

As varied as their musical outputs are, it’s interesting but not surprising that both John and Thiessen arrived at making music with samples through hip hop. As a musical tradition, hip hop provides a legibility to the potentials of sampling, as well as a rich culture where those traditions are practiced and continued. This connection feels especially compelling in the context of how use of generative AI is framed by its proponents as increasing access to people otherwise cut off or shut out from music production. The history of contemporary music is filled with people who started out with a cracked version of FL Studio, including (but not limited to) Metro Boomin, Kaytranada, Soulja Boy, and WondaGurl. In this light, claims that generative AI uniquely democratizes music production overlook the fact that artists have long accessed the field through informal, improvised, and often unauthorized pathways.

The labour of listening

John describes sampling as a search for the right detail — the right sonic fragment that can carry a feeling. While it can be tempting to adopt generative AI into his workflow to save on time and energy, his music is the result of his deep commitment to the work of listening, and through that, the slow accumulation of taste and perspective. “I put a lot of time into sourcing out the specific sounds and digging for stuff. I'm a music nerd anyways, so I always am finding and looking for stuff,” he says. John is explicit about where his samples come from: YouTube. For him, the internet is not a magic box that outputs finished songs. It is a vast, messy archive where you still have to hunt. “The fact that these things can just be minimized to [subscribing to] a website is just disgusting to me,” he reveals. For John, the act of listening and digging for new sounds is exactly what makes his musical choices as an artist meaningful, and there is an immense satisfaction that comes with aligning a sound with an experience. “Like, if I wrote about some relationship and I found a sound that represented it kind of perfectly, that’s the greatest joy to me,” John says. There is a joy in discovery, in recognition, in the moment the material answers back. “I never want to take away the joy of building collages,” he explains. 

Thiessen begins his process with physical digging, searching for rare materials — the weirder, the better. He characterizes his listening through these selections as a practice of split attention. “I’m always listening to new music and trying to find new sounds, and so you always kind of have one ear that’s listening just for enjoyable music. And then in the other ear it’s, ‘oh, there’s like a part where just the single instrument comes in, and would that make an interesting sample?’” He scours through thrift store selections or record store dollar bins, hoping to find something overlooked or discarded like a sound effect record that can provide his work with interesting rhythms or texture to build on. 

Both artists describe sampling as a practice where listening is work, and where that work is not an obstacle to creation, but the source of it. Meanwhile, music generation platforms like Suno depend on vast training datasets assembled through processes that are anything but frictionless. Behind the interface sit layers of invisible labour. Data annotation, content moderation, quality control, and filtering are frequently carried out by distributed workforces, often underpaid and concentrated in the Global South. The mythology of instant creation rests on coordinated human effort that rarely appears in marketing copy. 

It’s also unclear exactly what music these companies have trained their data on — Suno simply boasts their model was trained on “essentially all music files on the Internet,” revealing that it likely involved unlicensed music, further capitalizing on the hard work of others who did not consent to have their music used in such a way. Here, generative AI redistributes effort away from the visible practices of listening and construction, and toward the obscured economies that sustain the inflated venture capitalist dream. 

Sampling as memory and community

Sampling has another feature that becomes sharper in the era of instant generation. John talks about the balance between obscurity and recognizability, between reaching a broader audience and speaking to a local one. He wants his music to be broadly accessible to people who might not pick up on all of the sounds and texts he’s referencing, but also recognizes how sampling can create a kind of quiet link between fellow makers. “I reference it for [the community], the people around me who are doing similar approaches, who are into the same thing,” he says.

Thiessen describes that link with a vivid metaphor of recognition at a distance. “I know when I’m listening to music and I recognize a sample of an obscure song, I feel so connected to it all of a sudden.” He imagines it as shared discovery through shared channels, a sense that someone else traveled the same path through the archive. “You know this person obviously has gone through the same channels to discover this music and you all of a sudden feel a kinship,” he says. Sampling, in this view, is a kind of social technology and language.

That kind of lineage is markedly different when it comes to AI music tools. On user forums, people regularly describe generating near identical tracks across different prompts or across different days. One Suno user writes on Reddit, “[...] even with wildly different style prompts it basically gives me the same song with tiny variations.” Another complains about “an uncomfortable sameyness” when testing variations. On Udio’s subreddit, a user reports: “UDIO gave me the exact same song twice (over days).”

Those accounts don’t necessarily reveal how any model works internally, but the way the repetition emerges through AI music generation versus through sampling shows up as sameness without a shared source or relationship. If two users receive similar musical outputs, there is no rare YouTube find calcified with digital decay, no link to a dusty sound effects LP, and no drum break that you can trace through decades of songs. There is only the platform, and the platform’s tendency to converge for the sake of generating users’ prompts in the fastest way possible.

Sampling does the opposite. Even when a listener does not recognize a source, the history remains embedded. A sample is a trace of a recording, a room, a voice, a past moment. It is a quotation that offers community through shared materials, and it offers memory through audible residue that sticks around long after sounds are transformed.

How sampling informs creation

Zooming out from individual songs, it’s clear from both John and Thiessen’s work that sampling has enabled an omnivorous approach to their craft that empowers them to arrive at sounds that are uniquely their own. Throughout our conversation, John talks about immediate influences like MF Doom and Madlib’s classic Madvillainy album, but also the way that claire rousay incorporates field recordings into her tender songwriting as another significant source of inspiration. The first bridge of sand. record, 2023’s what now?, drew heavily from fellow Torontonian John Oswald, applying the plunderphonics principles Oswald established in the 1980s while transforming his sound sources more overtly to avoid the same litigious outcome. “I really was inspired by John Oswald's philosophy and trying to restructure sounds as a whole new thing. what now? is an attempt at a plunderphonics record with vocals on it,” he explains. Other songs draw more on ambient IDM and footwork, treating samples as percussive elements, with John pointing to producers like Actress and Jlin as significant influences. The beat on “Obsidian Shell” from his self-titled 2025 album tumbles forward like a crate of samples was just tipped over. 

Thiessen’s been at the sampling game for much longer, and the result is a career that is incredibly varied in its expression. There’s the early creative work he did as Bong Sample, imagining soundtracks for sci-fi thrillers yet to be made, the uneasy psychedelic collages he crafted on Pixel Geometry for his solo project Kaunsel, and the more avant approaches that he and Jacob Audrey Taves have been tapping into as New Chrome Mennonites, an entirely sample-based project.

Many of those projects are contained within his DIY tape label, Pseudo Laboratories—which he co-founded with Ian Rowley (Boothman, Home Front, Rhythm of Cruelty) and are featured alongside many more projects of Thiessen’s which involve electronics but not necessarily sampling. Lately, Thiessen’s been creatively invigorated by the practice, however, and sees it as a way to move fluidly between genres. “I've been doing a lot more sampling and I've been trying to go with this idea of essentially [being] able to play any genre of music,” he explains. “I could play a noise set or I could try to play at a hip hop show or even some kind of industrial metal thing. You could choose your samples based on the audience that you're going to be performing for. You have this ability to not break genres, but explore anywhere. Like you can go anywhere with that.”

Just as they can grab a psychedelic guitar riff and set it to a drum break from a funk record, the act of collage that becomes a production necessity when sampling across sounds, styles, and decades, it’s only natural for these producers to treat genre as something that is entirely mutable to their own ends. The way AI music generation platforms transmogrify entire histories of recorded sound often draws comparisons to sampling in practice, but it’s here where the relationships and histories detailed in the previous sections really play out. 

What remains when sound creation is instant?

Generative AI’s promise is seductive because it treats music as something anyone can access quickly, and because it wraps that access in the language of imagination and ownership. For people shut out of traditional music education, or for people who want to sketch ideas, that can feel liberating. But the more you look at that promise and how AI music generation delivers on it in practice, the more the lustre and allure falls away. The same promise also risks shrinking what music means, and even narrowing what is creatively possible, because it encourages a point of view where music is solely an output, one that arrives without a relationship to craft, creativity, labour, memory, or community.

Sampling keeps insisting on those relationships. It encourages a sense of lineage, whether through deliberate quotation or through the quieter kinship of shared digging. It treats music as constructed, situated, and communal. 

Thiessen has a blunt way of describing his view on AI music generation that speaks to just how disembodied it is from context and the choices and influences that actually make art interesting. “I always equate AI generation to telling someone about your dream,” he explains. “It’s only interesting to you. It’s like, ‘my brain thought this stuff without me.’ It's just this unconscious thing that you wake up to.”