tl;dr: 2021 has seen a flurry of new generative music NFT project launches, all with the potential to rethink traditional industry notions of celebrity, creativity, fan engagement and intellectual property. But they have yet to see the same consumer demand or financial upside as their immensely popular visual counterparts, for reasons that are equal parts technical, legal and cultural.
This is Part I of a five-part, collaborative research report that the Water & Music community has put together over the last two months on the state of music and Web3. Contributors to this research thread on generative music NFTs are listed at the bottom of this page, sorted by role. You can view the current state of our report rollout and a full list of our member-contributors by visiting stream.waterandmusic.com.
Perhaps one of the most compelling themes in the history of media and technology is, as Marshall McLuhan put it, “the medium is the message.” Namely, the way people express themselves — and, in turn, the way others perceive that expression — is directly shaped by the format in which that expression is delivered.
For better or for worse, this dynamic has defined the mainstream music industry every step of the way. A large reason why commercially released albums still typically last 40 to 50 minutes — even in the streaming era — is because a 12-inch, 33 RPM vinyl record could fit around 22 minutes of music per side, or 45 minutes in total. Similarly, songwriters now regularly tailor their song structures to get picked up by Spotify’s playlist algorithm, such as moving the song’s hook earlier in the track to reduce its chances of getting skipped and therefore deprioritized in Spotify’s playlist sequencing. The recent viral power of shortform apps like TikTok has further entrenched this mentality in the industry, skyrocketing the 15-second earworm as the primary unit of music creation. In all these cases, the distribution channel has a direct influence on the creative agenda.
If it is indeed true that the medium is the message, then what is the distinct musical language of Web3? Put another way, what kinds of creative processes does blockchain technology uniquely encourage and enable? What can we expect from crypto’s version of the 15-second hook?
Perhaps the clearest example of what Web3-native creativity looks like today comes from the visual world. Many of the most financially successful NFT projects of all time, from CryptoKitties and CryptoPunks in 2017 to Bored Ape Yacht Club and SquiggleDAO in 2021, all derive from a common creative template — namely, using code to generate hundreds or even thousands of unique artifacts that all share a common visual identity, with varying degrees of rarity that can be verified and monetized on-chain. Socially, people often refer to these projects as “PFPs” (which, depending on whom you ask, stands for “picture for proof” or “profile picture”), due to the surrounding culture of showcasing these NFTs on social media to signal one’s belonging to a given, exclusive collector base. These projects are also often generative in the sense of relying on code (on- or off-chain) to create a large number of visual assets in a short period of time, instead of a given artist making each asset by hand.
Slowly but surely, we’re seeing musicians adopt a similar mentality of using Web3 to experiment with large-scale, distributed creative processes for music and audio. Water & Music’s community has identified nearly 30 music-themed generative and PFP NFT projects that have launched this year — from Holly Herndon’s groundbreaking Holly+ voice synthesis project, to one-off generative audiovisual NFT collections like Invocation (Telefon Tel Aviv x EFFIXX) and Rituals (Justin Boreta x Aaron Penne), to more standard PFP-type projects like 6ix9ine’s TROLLz and Trippie Redd’s Trippie Headz. (Water & Music members can access the full list of generative/PFP music projects in a brand-new, members-only tab in our music/crypto dashboard.)
What makes these generative/PFP music projects especially exciting is not just their potential to push the boundaries of where fandom and creativity intersect, bringing previously fringe or niche ideas around crowdsourced creativity and distributed brand equity to the masses; it’s also the potential that Web3 has to facilitate more dynamic, innovative and sustainable business models around these kinds of creative projects that otherwise would not be financially sustainable.
That said, the reality of the situation is that the social and financial value of generative/PFP music NFT collections has not reached nearly the same level of enduring success as their visual counterparts. This may come as a surprise given that music has been a direct beneficiary of the modern Web3 wave in other ways (mostly through one-off NFT drops), and that music is not only an inherently cultural and social art form but is also increasingly visual in its modern expression.
The core questions we wanted to explore for this article are not only what shape Web3-native music PFP projects have been taking so far this year, but also the challenges that remain to scaling these projects to mainstream appeal. Through our own market research and interviews with several founders behind generative music NFT projects, we found that closing this market gap is definitely not going to be as simple as just copying-and-pasting common practices from the visual art world into the music world; in fact, it may require rethinking entire traditional music-industry notions of celebrity, intellectual property and fan engagement from the ground up. While all the technical requirements to make generative/PFP music NFTs work exist, the social, cultural and legal foundations arguably do not.
We’ll approach this argument by exploring the following issues in order:
- A bit more context on the history of generative music creation, why it matters for music and why it makes sense that generative music artists are adopting Web3;
- Analyzing the technical plausibility of generative/PFP music NFT projects — specifically, how well certain tactics used in visual PFP projects like Bored Ape and CryptoPunks map onto musical collections, and
- Diving into the social, cultural and legal challenges to scaling generative music NFT projects, and the unique experimental opportunities that those challenges offer to artists.
Creative context: Generative music composers as “gardeners”
Renowned artist and producer Brian Eno coined the term “generative music” in the 1990s to describe a systems-based paradigm for music creation, whereby the composer builds some kind of automaton (i.e. pre-programmed, self-operating system) that generates music on its own, instead of the humans themselves working directly with sound to create a discrete composition. Eno drew his inspiration from contemporary scientific ideas at the time, including cybernetics, nonlinear systems theory and chaos theory — the key shared insight of which is that even the simplest systems can generate complex behaviors. A set of rules, tuned just right, can have creative potential that is deep and unforeseeable.
Inspired by these ideas, Eno reported a radical shift in his thinking in the early 1970s that changed his understanding of the musical composer from an “architect” to a “gardener.” As he shared in a 2011 interview, this is “the idea of the composer from somebody who stood at the top of a process and dictated precisely how it was carried out, to somebody who stood at the bottom of a process who carefully planted rather well-selected seeds, hopefully, and watched them turn into something.” In other words, the composer, under this new paradigm, is in the business of providing merely the ingredients for creativity, rather than a finished product or even a clear set of instructions.
Importantly, generative music does not have to use computers. Eno’s breakthrough ambient album Discreet Music uses asynchronous magnetic tape loops to create endlessly novel conjunctions of pre-recorded material; Steve Reich uses similar means to staggering effect in It’s Gonna Rain. But computers are obviously an ideal medium for designing generative automata, because as any coder can tell you, the fruits of programming are all about abstraction — and creating music by creating a system that makes music is simply composing on a higher level of abstraction. More recently, apps like Endel, Boomy and Jean-Michel Jarre’s EōN app have attempted to bring these computerized generative and adaptive music experiences to a more mainstream fan base, often raising millions of dollars in venture-capital funding in the process.
In Web3 communities, the word “generative” is applied broadly to art projects that use code to make large numbers of unique objects on a blockchain. The simplest form of a “generative” blockchain artwork might be the PFP-forward format that mixes and matches different features onto a cartoon avatar, as with BAYC. But in these cases, the algorithm simply compiles limited combinations of pre-made elements, which barely satisfies the definition of “generative” that artists like Eno originally intended. On the other end of the spectrum are collections like deafbeef and Tyler Hobbs’ Fidenzas, in which both the results and the code that creates them are treated as stunning works of art.
Challenges still remain with making the creative output of music generation more legible and palatable to a more mainstream audience. For one, even on-chain creative code is often still rendered in-browser by off-chain libraries (e.g. p5.js for Audioglyphs and Art Blocks), and the audio output of these libraries still sounds rudimentary compared to the high-quality studio productions that music fans are used to consuming, let alone paying for.
Nonetheless, in the context of blockchain and NFTs, generative approaches to art creation will continue to fill in market gaps as long as collectors want more unique pieces than creators have time to craft by hand. In this regard, it might be similar to other methods of cheap manufacturing — except it transcends mass production to the extent that the generative algorithm is itself treated as the art.
What makes Web3 especially compelling as a potential distribution layer on top of generative music experiences is the ability to make both specific musical outputs and their formulas collectable — as, with most of these projects, both the underlying generation logic/script and the end result are stored on-chain. This aligns well with a major collector motivation around generative NFT projects in general: A given on-chain art piece feels more permanent if it comes with the code required to render it.
Base mechanics of generative music NFTs: Rarity, blind mints and burns
If the goal of certain generative music NFT projects is to recreate the financial success of their visual counterparts, we can start by analyzing what traits and elements from visual PFP projects translate well onto musical ones.
On a technical level, a lot of this translation is plausible, and already being done. For example, one common element in visual PFP projects involves assigning a relative level of rarity to different visual or character elements (e.g. clothing, hair color) that might get incorporated into the final artworks. Rarity is a crucial factor in how crypto enthusiasts compare and price their NFTs relative to each other, and there’s even an entire cottage industry around helping crypto brands generate rarity tables for their NFT collections. Hence, many generative music NFT projects like SoundArts (pictured above) already implement a rarity-based approach, whereby developers assign a rarity level to each stem that informs the generative creation process for a given music NFT collection.
Another implementation element that many generative music NFT projects borrow from the visual world is blind minting, a process by which collectors make a primary purchase of an NFT without knowing the exact combination or rarity of the traits they will get. For instance, artist Julian Mudd’s Muddy collection allows collectors to mint a total of 1,000 generative NFTs around his song “Growing Pains,” each of which has one of 10,000 possible unique derivative combinations of vocals, chords, fills, bass and percussion around the song that the collector does not know full details about until after they make their purchase. (The static, original version of “Growing Pains” is available on Spotify and other streaming services.)
Projects with blind minting typically trigger a “reveal” of their underlying rarity info after all the NFTs are minted or once the minting period is closed, whichever comes first. Hence, by design, the process of going through a blind mint is strikingly similar to that of purchasing a loot box in a video game, or a lottery ticket IRL.
“I think that there is a strange excitement about not knowing what you are going to get when minting,” Patrick Price (aka Patty G), who started 3Q Collectibles, an upcoming project that collaborates with producers to create foundational stems for generative audio NFT collections with built-in rarity rankings, tells us in an interview. “The feeling of, ‘we are all minting at the same price, but I may draw something that will be worth so much more’ … taps into the same psychology that makes us love gambling.”
A third mechanic we’re starting to see emerge in the generative music NFT landscape is giving collectors the ability to burn tokens (i.e. remove them entirely and irreversibly from supply), creating additional scarcity in the “genesis” collection as the number of original assets decreases. As a sort of performance art piece, digital artist Pak launched a dedicated website burn.art where any collector could burn any NFT they owned in exchange for earning the token $ASH, which could be redeemed further down the line for Pak’s own NFT collections.
In the music world, flagship generative music NFT project EulerBeats structured their first drop as 27 one-of-one genesis “LPs”; for each genesis LP, 120 total copies or “prints” were sold on a bonding curve (i.e. automatically increasing in price with every sale), with 8% of print sales going to the genesis LP holder, 2% to EulerBeats and the remaining 90% to a burn reserve treasury, according to the project’s terms of service. Further gamification architecture presented NFT owners with the opportunity to burn their original print, in exchange for 90% of the print’s current value.
Social, cultural and legal challenges
While all of the above tactics are technically feasible for generative/PFP music NFT projects, whether the paradigm is socially, culturally and legally conceivable at scale is another question entirely.
Evaluating the potential market for generative/PFP music NFT collections starts with not just a financial exercise, but a philosophical one: Why do people even enjoy listening to and sharing music in the first place? As an inherently social art form, music arguably accrues additional value the more it is shared — so does everyone necessarily want to own a unique piece of music, or do they all want to collectively share and experience the same one? Does a fan’s relationship with music change once you are seemingly able to mass-produce multiple different versions of that music with the push of a button (even if the artist may have put in many more hours of hard work behind the scenes to set up that system)? And how do traditional notions of top-down celebrity branding — where fans follow artists for having a distinct personality and sound that cannot be replicated — map neatly onto decentralized infrastructure?
Lack of social utility
In general, one of the most powerful psychological draws of generative/PFP projects is the ability to unite collectors under some kind of umbrella community. “Generative PFPs are the perfect example of having commonality, but also uniqueness,” SoundArts founder Paris Blohm tells us in an interview. Citing BAYC as an example, SoundArts founding member Brian Nguyen adds that “we are all Apes, but we all have our own identity within that community.”
Many generative music NFT projects in our database lean into a similar, built-in community element, usually by treating NFT ownership as a path to access and governance in a collection-specific DAO (Holly+, Mudd DAO, BeetsDAO and BleepsDAO being a handful of key examples). “We see generative NFTs potentially creating a new kind of fan club, where NFT holders can access future drops and shows from an artist,” Kalam Ali, co-founder of experimental music/art ventures So Lab X, Sound Obsessed and IN X SPACE, tells us in an interview. “We have this new way of aggregating a fan base around something they own, in the form of a larger-scale NFT series from an artist or band. We also see the format’s creative and financial potential going hand-in-hand, such as using generative NFTs to replace merch or even tickets in virtual metaverse events.”
That said, these community experiences are still built largely by and for a highly web3-native artist and collector base. The mainstream social utility of generative/PFP music NFTs still falls far short of that of their visual counterparts, for a variety of cultural and technical reasons.
For one, music is inherently more difficult to browse than visual art. Over the course of listening to one song, a given fan or collector can not only scan hundreds of visual PFPs, but can much more rapidly analyze the rarity and distinctiveness of these visual assets in comparison to others in the same collection. In contrast, at least for the everyday consumer, assessing the rarity of a given audio file (e.g. deducing slight variations between different stems) may not be as intuitive.
In addition, a large part of the culture around visual generative/PFP NFTs is the ability to self-display one’s NFT as a profile picture on Twitter, Discord, and other social platforms, especially for collections like BAYC and CryptoPunks that are designed around human or humanoid characters. For the most part, no mainstream social platform supports music or audio as a format for PFPs.
These limitations also raise questions about which parts of generative music NFTs owners will ultimately identify with more — the audio, the visual, the subsequent DAO or perhaps the artist themselves behind the project? One potential path forward for scaling generative/PFP music NFTs to a wider audience may involve giving them a humanoid visual layer on top to make them more “relatable” to everyday fans, especially to those who may already be familiar with the BAYCs of the world. For instance, the interactive, generative music NFT project WarpSound (pictured above) represents its musical output visually via a collective of virtual, human/anthropomorphic DJs, unlike many other similar projects whose visual layer consists solely of more abstract art.
Lack of legal recourse
The visual PFP world would not be as successful as it is today were it not for its open embrace of “remix” culture — namely, giving NFT owners at least some commercial rights to commercialize their tokens. One of the most notable examples of this dynamic can be found in the BAYC community, whereby owners have full commercial rights over whatever NFTs they own from the collection. Nowadays, you can find Apes on T-shirts, coffee mugs, hats, comic strips and countless other kinds of products, not to mention in a whole new supergroup under Universal Music Group; this openness to derivative IP enterprises has helped to catapult BAYC into one of the most financially successful NFT collections of all time.
(Importantly, not all PFP projects adopt this mentality; the creators of CryptoPunks, for example, retain the sole right to profit off the project commercially, while CryptoKitty owners can earn only $100,000 a year in sales from their own commercialized versions of their artwork. But even this right is somewhat loosely enforced, and more limited projects tend not to generate as much revenue in the wider community. It seems that most PFP projects have decided that brand expansion and awareness is more important than protecting the IP.)
There are some early, small-scale proofs-of-concept of this hands-off approach to derivative works with music NFTs. For instance, the Async platform enables collectors to own NFTs representing the right to change a given stem in the final master of a song at any given time, and then to purchase NFT “prints” of different combinations of these stems (similar to EulerBeats’ genesis vs. print model described earlier). Projects like Audioglyphs and EulerBeats also give original NFT holders commercial rights to the associated songs for as long as they hold the original tokens.
Such a clause, though, usually relies heavily on trust and the inherent assumption that collectors in the community are not bad actors; actually enforcing this link between IP exploitation rights and token ownership is done rather loosely, if it’s even done at all. For this reason, it will be an immense legal challenge to scale derivative monetization models around generative music NFTs to the level of mainstream music culture, especially in a way that meshes well with traditional music rights holders.
In fact, even before thinking about Web3 and NFTs, the legal issues around generative music creation are, to put it lightly, a shitshow. There is no legal standard in most countries for whom exactly to credit in a songwriting or production process that incorporates AI — namely, whether it should be the enabling software, the human producer of the original stems or source material or some combination of the two who should be the “rightful” copyright owners behind a piece of AI-generated music. For generative music NFTs specifically, tokens only amplify rather than remove entrenched legal complexities in the music industry — a trend that is perhaps ignored by the strong anti-copyright ethos in the Web3 community, given the extent to which PFP projects in general freely lift IP from each other with no legal recourse.
Conclusion: Creativity as its own (financial) reward
The bottom line is that for a variety of reasons, consumer/fan demand — and therefore financial models and value — around generative music NFTs are not yet proven. Yet, at least for the time being, these same limitations serve as a shield from speculation, at a time when much of the music/Web3 ecosystem is concerned with building infrastructure with artists’ wellbeing at the core.
Visual PFP projects like BAYC and CryptoPunks have been successful not only because they have built a sense of shared community and identity, but also because they have become tradable assets with some financial utility, which in some cases has led to highly speculative markets for digital art. But the inherent characteristics of music and the function it plays in society arguably make it unlikely to follow that same speculative route.
So, as a closing point, now is an opportune time to think about what new kinds of business models can support the next wave of generative music NFTs, separate from any expectation of financial speculation.
Perhaps “flipping” NFTs is the wrong model for artists in this ecosystem (or for any artist in Web3 for the long term, for that matter). Instead, generative music NFT projects could focus more on building long-term models for earning passive income for the original artist(s), while also empowering collectors to build their own businesses and creative works around the tokens. Holly+ has taken this approach with their own auction house on Zora (screenshot above), where anyone can submit artwork made using the Holly+ voice model for potential inclusion in the Holly+ DAO’s 1/1 NFT collection on the platform. Profits generated from sales of these crowdsourced NFTs are split 50% with the contributing artist, with 40% going to the Holly+ DAO Treasury to fund new tools, and 10% going to Holly Herndon as compensation for the use of her digital likeness.
There’s also a potentially interesting financial opportunity in catering generative music NFT projects to artists and the music industry, not just to fans, by monetizing the underlying creative process alongside any discrete creative output. “We’re interested in looking at ways to use NFTs and tokens to monetize generative music workflows, templates and datasets,” says Ali, referencing open-source, off-chain resources for aggregating generative art models like Pollinations.Ai as inspiration. “You could use tokens to get access to certain art datasets, or you could buy an artist’s NFT to obtain a license to the creative software used to generate that artwork.”
Such a monetization model harks back to Eno’s notion of the composer as a bottom-up gardener, versus the composer as a top-down architect — and how this distinction reframes Web3 artistry as providing not just finished creative products, but rather the seeds for NFT collectors and other artists to create and monetize their own derivative works.
“We kind of see this as the NFT format pushing art forward in the same way that the development of recording technology created the idea of the recording artist, where the actual recording of the music is the art as much as the notes or lyrics of a song,” the team behind Synthopia, a generative music NFT project launched by Gramatik, Luxas and the Audioglyphs platform, tells us in a statement. “If NFTs make the economics of generative music work, we think it can have the same kind of impact.”
From the artist’s perspective, this may be the ultimate unlock for Web3 in the music industry in the long term — thinking not just about the short-term financial opportunity, but also about the deeper frontiers that blockchain and decentralization open up for artistry at large, for which any financial upside is just the icing on the cake.
This article is free for the wider public to consume and share. To view the full list we’ve compiled of nearly 30 generative music NFT projects that have launched this year, you can navigate to the “Generative/PFP Music Projects” tab in our members-only Music Crypto Dashboard.
Cherie Hu (A, B, C, D)
Yung Spielburg (A, B, C, D)
Elliot Cole (B, D)
Andres Botero (B, D)
Jillian Jones (B)
Jonathan Larson (D)
Levi Downey (D)
Ana Carolina Laurindo (E)
Kalam Ali (F)
Patrick Price (F)
Jeff Nicholas (F)
Brady Keehn a.k.a. Panther Modern (F)
(A) Research project leads
(D) Data/project curators
(F) Member interview sources