Spleetercore: The World’s First Album Fully produced by AI
Meet the world’s first album fully produced with samples extracted by Artificial Intelligence.
Meet the world’s first album fully produced with samples extracted by Artificial Intelligence.
Moises “was born” at the end of 2019 and already gathered more than half a million passionate users worldwide. After being adopted by karaoke enthusiasts, music teachers, and students, Moises also became the platform of choice for those who want to remix and produce new music.
An example of this is a project by the band X-spleet /// Spleetercore, from the U.K. They created an entire album with samples extracted by Moises’ Artificial Intelligence. We talked with the creator of X-spleet /// Spleetercore, to learn more about the band and the album. Check it out!
MT: Why this name, and how did you come up with it?
This is music made using sounds revealed by Spleeter. I was torn between Spleetercore and X-spleet for the music and a pseudonym, and I couldn't decide. So I'm using both - for both.
MT: We saw some contributors in your volume one. How would you describe your music initiative? It is a group or solo artist?
First and foremost, I am a music maker, and I've been into sampling for years, so when I found out about Spleeter/Moises I couldn't wait to crack open as many songs as possible and start using the stems to play with.
I have also recently started to train as a teacher and teach a few music classes at a college. The contributors to Spleetercore volume one are all students who enthusiastically joined in with the experiment to separate songs through Moises and then put them back together in the wrong order.
MT: How do you collaborate and how this project started?
I asked the students to select 5 songs each and then separate them through Moises/Spleeter and create a new piece by using a track from each of the 5 songs, for example, using a Beyonce vocal track with Drums from a Foo Fighters song and Bass from a David Bowie song.
The next stage was to resist the urge to make all of these tracks fit together by changing the pitch and the tempo, etc., and just let them play together - to hear what that sounded like, and obviously, it seems insane!
The reaction to hearing this stuff for the first time was exhilarating and disorientating. I really felt very strange afterward. The students loved it (I think mainly because they are teenagers and most teenagers like absurd gross things). I considered abandoning the project because I thought we might have done some kind of irreparable damage to the universe, but then I checked, and the universe was fine, so we persevered.
I do think this kind of misuse was an important first port of call for the new technology. Someone had to do it.
MT: What first got you into music, tell us more about your background (which instruments you play, etc.)?
I first got into music because I wanted to be in a band - I taught myself for a bit, but then I studied music at college and university. I play traditional instruments and music, but I prefer using a computer to do most of it now. It means I can eat at the same time.
MT: Who inspired you to make music, any artist, or relative?
I think Nirvana was the 1st band that made me want to play music. I've paid tribute on Spleetercore 2 by doing a remix of 'Hairspray Queen'. It's called Digital Hairspray, and it's not as good as the original. I just wanted Kurt Cobain's voice from that specific song, which I've hidden in other compositions.
MT: What is your creative process like, maybe share what you mean by applying a sonic roulette?
I think I'm only really interested in novelty. I want to hear things that I haven't heard before, so I'll use any process to try and find that. The first Spleetercore experiments are just one way to use this algorithm to find new sounds and make new music. I'm sure we'll all find other ways of abusing it, but for now, I love the sound of all those disparate tracks all playing at once and fighting for your attention, it's a bit like free Jazz.
The sonic roulette idea is just an extension of that but without me going to the effort of breaking all these pieces of music apart and putting them back together in the wrong order. Instead, why not have an app that's connected to a database of music that can just do it for me, selecting ‘pre spleetered’ stems to all play at the same time. Frankenstein music. Maybe you could incorporate this into a Moises update…? I'm sure I'm not the only weirdo who would love to have this happen.
MT: How would you describe the music that you create?
Some spleetercore should be unsettling and strange and new. Some spleetercore should be beautiful and haunting and unique. Some spleetercore should make the impossible possible.
At the moment, though, in its infancy, it can simply be described as a bit short and a bit weird, like Tom Cruise.
MT: Do you feel your music fits into any genre, you mentioned experimental/noise, how would you describe your music?
I want it to be its own genre, with the only rule being that you make new things from the dismantled parts of other music—no other instruments allowed.
It's recycling, isn't it, or upcycling or more likely it's downcycling or its pop finally eating itself (...).
MT: Who would you most like to collaborate with?
I’d like the Spleetercore genre to continue to grow, so I would like to collaborate with anyone who wants to make new music using the dismantled parts of other people's music.
I'd like Spleetercore to be a label, and a hub for the new wave of A.I. assisted music.
MT: It looks like you're leading a new generation of songwriting and creative music in general. How do you envision the future of music post-AI?
I think it's going to change a lot of things, but I don't think musicians should be worried, it's not going to replace them or their methods totally. You're never going to get A.I. to replicate a disappointing version of 'sweet home Alabama' played in a bar to 3 people who aren't listening. That is strictly the business of human musicians with human dreams, and they should not be infected by Artificial intelligence or any other kind of intelligence.
MT: What should we expect in the next few years, and how this might impact the music industry in your view?
The jukebox project looks like a fascinating application. Having any artist sing any song in any style. Though it will probably feed right into the eternal death parade, that is the touring hologram industry. So when you watch the Amy Winehouse hologram perform, it'll be able to take requests. Aside from that, I don't want to share any of my ideas publicly until I've spoken to Elon and Grimes.
MT: It looks like AI Mastering is on the rise. Do you believe one day we will also have AI Mixing? What would it look like?
I'm not sure any of it is a good idea. Mixing and mastering are artforms in their own right, and it's not something you should pass through some kind of instantaneous victory machine. We should concentrate on making A.I. bring about new forms and new ways of interacting/playing and listening rather than making mix engineers redundant.
MT: How would you describe to the broader audience what Moises.ai does and enables?
I would describe it as an easy to use interface that will bring about a new wave of sampled/plunderphonic music. I think that's a big deal.
It would be nice to see some extra features added like pitching and looping for each track, and then if you add that to a 'mix and match' aspect from a database, you've effectively created a DAW… It will also help drunk people to do Karaoke on public transport.
MT: How do you feel the AI has impacted the music business?
Well, at the moment, it's mostly helping curate playlists and recommendations for people. I think that was also a job we could have left for the humans - D.J.'s, labels, writers, and friends are always better curators, and I miss them. I think the way it's used at the moment is driven by corporate demand for data. If you have access to people's listening habits, you can try to use that information to sell them a phone or shower curtains or other bollocks. I might be being overly cynical, but I think it's true. It's utterly boring isn't it - It's nice that Deezer had put this out as open-source - to see what people can do with it. I hope we can play a small part in realizing its potential as a creative force rather than another dull attempt to make more money or more data for money or more money for data.
MT: If you could change anything about the industry, what would it be?
I would set a time limit on the availability of songs on streaming services. After a month, you no longer get the song to keep in your playlist. Instead, the streaming service has to provide a link where you pay the artist directly for their music. This would obviously only work for artists who are alive, but they're the ones who might need the cash - Streaming services would still make some money, but they would just be hosting and distributing rather than pretending that all the music belongs to them.
MT: Looks like you're releasing volume two soon, how would you describe the music you make and what should we expect in comparison to volume 1?
Spleetercore 2 is a lot less harrowing than volume 1. It's a lot more musical. It's like an A.i. search engine D.J. set, scanning quickly between a muddle of mishaps, bad connections, and strange loops.
I think it's a decent demonstration of what's possible with spleeter/Moises, i.e. combining parts of music from different genres and different eras. This time I tried a few different techniques rather than just putting things together that will obviously sound strange ( although this did still happen). I tried putting things together that might just work with a bit of tweaking. I tried fitting songs together that might have a had a lyric in common (like RESPECT by Aretha and Walk by Pantera on track 8). I tried putting the vocals from the original with the music of a cover version(track 3), there are lots of other musical connections as well, but they might only mean something to me. It's still a tiny investigation into the potential that's been opened up by this software, and I'm just scratching the surface.
MT: Who did you collaborate with for the second volume, and how did these people contribute to the final work?
Due to lockdown, I did this all alone. And also, getting the students to join in with this was probably a one-off, I should perhaps try to teach them something useful.
MT: Can you share more about your Mixing and Mastering process for volume 2?
I used Ableton to tune, loop, arrange, and time-stretch the samples, and I used a few VST effects like reverb, echo, and EQ. I think many producers refuse to work with the spleeter output at the moment because of the audible artifacts, but I really like the objects - it's like ghost music. Entrails. I made a few tracks just using the artifacts and the other interesting effects created by the algorithms removal/rewriting (whatever it is that it's doing, I'm still too stupid to understand fully).
MT: Did you use any instruments on volume 2? Which tools did you use?
No instruments, no VST, just Moises and Ableton.
MT: Will you ever perform live or have you already?
I might if someone can make a hologram of me.
What’s next for you, any plans for volume 3?
Yes, Spleetercore 3 is underway. It's using the work of 4 artists specifically. They all have something in common. I won't reveal the secret just yet
To check out the album and accompany X-Spleet /// Sleetercore: