Skip to main content

Holly Herndon: "Computers surprise you in a way that an acoustic instrument doesn’t"

Holly Herndon
(Image credit: press)

It’s not really doing down other computer-based musicians to suggest that Berlin-based, Tennessee-born Holly Herndon is a computer musician par excellence.

With a CV to make you weep, this respected Stanford academic of all things musical probably didn’t need to bother topping it off by making one of the most critically lauded, technically complex – and somehow deeply human – albums of the past few years. But she went ahead and did it anyway. 

Herndon is quick to point out that her most recent album, Proto, is actually the work of a ragtag cast of characters akin to something out of a sci-fi novel: chiefly herself, collaborators Mat Dryhurst, AI expert Jules LaPlace… and a personified AI programme named Spawn, aka a DIY, neural network-based piece of software housed in a “souped-up gaming PC”.

In fact, Proto’s background is so rich – and endlessly fascinating – that we concur that it fully deserved its own podcast. We caught up with Herndon mid-lockdown to hear more…

Holly Herndon

(Image credit: press)

Firstly, given the circumstances, how are you doing?

“Well, I’m used to working from home, in Berlin, but of course our entire concert season has been obliterated which is very sad.”

Has it affected other plans too?

“It’s been majorly disruptive. Any plans to go anywhere and do site-specific work. But like I say, I’m fortunate enough to be able to accomplish a lot from my studio.

"My partner and I have actually been wanting to start a podcast for a long time [available now on Patreon]. Because so much of what we do is research-based, the album is just one piece of a much bigger project - the music industry-facing side. 

"Conversations are sometimes the best way. We were always travelling too much to put it together. This has forced us into doing something we’ve been wanting to do a long time. So it’s a silver lining!”

There is definitely so much extra to be said about what you do behind the scenes!

“It’s really hard with music because it’s an abstract language. It’s not writing an essay. There is so much we want to share. There are so many smart and interesting researchers who we come across. We feel like a musical audience may not be familiar with them. I don’t see the ideation as something that happens in isolation. It’s not about a lone genius on a mountain top. I don’t buy into that a lot.”

How does one of your projects get started? They are so vast in scope, so we appreciate that that’s a big question!

“There’s a lot of stumbling in the dark. It’s a process of whittling away at it. I always return to a community of thinkers for different takes. It’s a matter of trying not to accept the surface argument.

"When it came to machine learning it was about creating a counter narrative. How is our way of thinking about technology framed by our history? It’s a matter of researching and whittling away. It’s often a cloud of references and thought processes that becomes a piece of art. It can be very frustrating. Almost like hell at times actually! [laughs]”

Coming from an academic background, is it a bit like doing a PhD?

“Yes, it’s about trying to contribute something to a conversation, rather than doing something that’s already been done. And hopefully coming up with some original thoughts.

"Making our own tools has been a big part of the research. What that says about the person making the tool. They come from somewhere. Having some agency in the technology you use is a big part of my work. Platform was already quite collaborative but Proto was on a new level.

"The challenges were therefore logistical. It was time-consuming in a boring way. I’m so used to digital processing being immediate. As a musician I don’t have to depend on rendering as much. But dealing with Spawn and machine-learning, you’re dealing with long waits for output. And you have to readjust what your input is. It was really laborious.

"It made me really appreciate back in the old days when Stockhausen had his punchcards… but even before then. Computer musicians would have had to be dealing with different timescales.”

People talk about computer music-making being so easy these days, but in some ways you seem to have run in the opposite direction! Was the process fun?

“The thing about computers is that they surprise you in a way that an acoustic instrument doesn’t. There was a six-month period where the output wasn’t actually interesting. But we had breakthrough moments which made up for it. I’m always looking for sounds outside of nature. But the new timbral information you get from it… that’s what’s really exciting to me.

"I always think of how it must have been for John Chowning who is a friend of mine [one of the founders of CCRMA, where Holly was studying at Stamford - the Center for Computer Research in Music and Acoustics]. He discovered atom synthesis, tweaking oscillators, realising that when you manipulate one oscillator it would change the pitch and timbral information of the other. And when the two are combined you can manipulate that.

"With his composer’s ear he understood the implications of that. There must have been so many boring hours of non-interesting oscillator fiddling and then he came across this discovery that had so much impact on the field. Not trying to make a comparison with myself in that way, but that’s the tradition I’m coming from. I find that kind of stuff really exciting.”

“A huge point to music is a sense of connection… anything worth doing is joyful and laborious”

Talking about foley sounds, did you have any fun moments with that on Proto?

“We did allow ourselves to have fun! One of the reasons I wanted to put together a vocal ensemble was because I was feeling lonely in the studio and I wanted to feel that sense of ‘musicing’ with other people in real time. So that’s where this interesting parallel came about with neural networks. I started to think of them as one of the most contemporary ways for humans to coordinate. And vocal ensembles were one of the earliest ways that humans found to coordinate, whether it was for hunting or rituals.

"I started looking at this desire for human coordination across this really long time horizon, coordinating vocal techniques with AI. And, of course, it’s really fun to sing with other people. When I was young I used to enjoy being in choirs. It’s a very joyful thing. Especially as a computer musician who spends hours in the studio.

"This is such a dorky thing but even when I was sitting editing for this one, I would hear some of the little jokes the group were telling between takes and sit there laughing! It just made it all that bit more joyous. Because ultimately, a huge point to music is the sense of connection and communion! I think anything that’s worth doing is both joyful and laborious.” 

Proto does seem quite a human album…

“There’s this weird perception of me as a machine computer girl, but computers are part of us! It makes total sense that it sounds human because it is. It’s part of us. I think when we start to see AI - well, actually, I prefer the phrase machine learning, it’s more descriptive - as a separate entity, we erase the human input that went into it. We’re starting to see this as some kind of alien other that came out of nowhere but that’s misleading. So it is very human.”

How do you transfer your work to live shows? Has it posed technical challenges?

“I think any time you translate music from a studio to a stage it’s different. But since my first album, Movement, I was playing the songs on it live before I even considered recording them.

"People said, ‘you should record this’, but I thought it was too weird! I didn’t think anyone would care about it. And then it was like, wow, people are into this. So that’s remained a hallmark of my practices.

"Of course, SPAWN required a lot with the rendering, but otherwise so much of it is about realtime vocal processing - so we were performing it in the studio to perform it already. That translation, for me, is not as alien to me as for some others. And for me it’s really fun.

"I love taking an album and making it translate to a stage. Especially if it’s a really nice stage! We played the Barbican which I love, with its acoustics, everyone being so nice and the whole plushness. One of the stage guys there suggested we use a scrim, which is a kind of sheer projection screen. We put lights behind it and put two of the singers on that. It transformed the space into a different world. And I think in the day people had that kind of thing in a church, but a concert can provide a secular version of that! There’s something therapeutic about being emotional in public with other people.” 

You’ve said ‘a naked voice is terrifying to you’? Where does that come from?

“Before using the computer as my primary instrument I don’t think I had come into my own, in a way. Previously I hadn’t found anything that sounded like me. I was always emulating.

"It was a huge revelation as a composer. I was trying to find a way to make it an embodied performance. Before that, people were complaining that it looked like you were checking your email. I think that’s not quite fair - I personally love the sonic properties of being out and listening on a great sound system, but something was missing for me. It’s difficult for the audience to understand the intent of the performer if there’s nothing happening necessarily.

"The mirror neurons in the brains of the audience actually start firing when they see the emotion of the performer. So I started using my voice and realtime digital processing as a way to embody my performance with the computer. I never saw myself as a vocalist. I was always just interested in the voice as an input source – not a concert instrument.

"So that’s the origin of that idea! I was looking for ways to transcend the limitations of my body through this incredible toolset. I was able to make a voice model of myself and get Spawn to sing things that I wouldn’t be able to. The culmination of that was the track GodMother with J-Lin. That’s my voice model performing JLin’s rhythm. Spawn is already outperforming me.

"Anyway, vocal processing has a longer history. Think about people in the 13th century. A giant church is just a huge reverb unit. That aesthetic is still with us. The idealised version of a human voice.”

“It’s difficult for a live audience to understand the performer if there’s nothing happening”

Do you think the culture surrounding performance setup has changed?

“I’ve seen a huge shift in this. A lot of people perform with laptops for playback and the rest is performance art. Sometimes you’re super engaging, but sometimes you come across as lazy. Maybe I’m a masochist but I like realtime processing. It’s a living thing.”

Tell us a bit about your workspace…

“Depending on what I’m working on, I have commercial software as well as niche ones from research institutions. And some I’ve made myself. Whatever I’m sketching I’ll go to what I trust. But I’ll always have a microphone plugged in ready.

"When I’m using Spawn, that’s basically recording and uploading files. I have one of the longer MIDI keyboards and Adam 6 horizontal speakers with the tweeters. They are great. A very full, clear sound. They’re based in Berlin so I was able to go to their warehouse and play some of my music on different setups. They also have a really cute section where all of the tweeters are hand-folded. There were lots of old ladies doing it. It felt like Santa’s workshop or something.

"I also have a Focusrite Clarett soundcard, with low latency, that I appreciate. I use some from their Red pro line when I’m touring. I had never used D sub-cables before, but they’re a great way to reduce your footprint if you’re using lots of cables. The pins are so much smaller, like a 1/4-inch pin. So it’s pretty simple: that, my laptop and a crappy mixer.”

Any hardware?

“I haven’t opened up the modular can of worms yet. My partner and I used to make more jammy dance music, but it wasn’t really for release, more just for parties. We had more hardware for that - a 303, 808, 707, Drumfire and all that. That stuff’s always really fun but moving continents (from California to Germany) has made us shed a lot of it. But sometimes I do crave that tactile experience. Who knows, maybe for the next album. I like to MIDI-map things.

"One thing is to have a really well soundproofed room. I got a builder in - Ben - who did a really beautiful job. The room strangely feels bigger because it was handmade with great materials - even though it’s actually made it smaller. It’s also lit in orange coloured lights, which makes it feel cosy!

"We have a Barcus Berry contact mic that’s probably more of a thing that you’d hear on Platform, but I always create my own sample base and do my own foley to smash together a palette. Sony PCM is a really great portable field recorder that I use all the time. That’s basically it.”

Can you tell us how you got into the kind of projects you’ve been involved in? Short of just filling in a university application! 

“There are plenty of artists who work in an ‘extra-musical’ space who aren’t academic. So I wouldn’t want to pigeonhole the kind of work I do as being only possible in an academic space. That was my path, but a lot of contemporaries didn’t. But for me I’m not the kind of person who just jams. I guess I need a starting point or conceptual motivation to get things started.

"I think that comes from a need to have my own voice… because music is a shared language so we’re never able to break completely free from our musical history. For me it was important that I find a new framework. It doesn’t have to be a grand vision, but instead just a new process chain. A lot of it was that I didn’t want to use the presets available to me. It felt like it was making all the female voices ephemeral and beautiful, and that’s not what I wanted. To do that I felt like I needed to build my own patches. Trying to set up something that felt unique to me.

"Of course I’m coming from an academic background so I’m being asked to justify my decisions. But I like that! Analyse or be analysed! But that’s not for everyone and I don’t think people need to do that to be taken seriously as a musician. For me it was about avoiding getting stuck on trying to sound like someone else. That’s just boring!”

What themes are in your new stuff?

“Nothing I would really talk about right now, but I’m working on two simultaneous projects with two different feelings to them. So if I’m in a hyped-up mood, I go with one and then if I’m in a more reflective mood the other. I’m trying to use quarantine time to be somewhat productive. But also understanding that it’s fucking traumatic and scary, watching this disease ravage the world. I’m really far away from my family, so watching it all on TV makes it a sad time. It’s OK to mourn and not try to be a hyper-productive machine.

"I miss touring. My attitude to that has changed. I can’t wait to play again. I think I’m going to get over, for example, annoying soundcheck issues more quickly, in the future, because I’ll appreciate it more!”

“Nobody needs me to record an acoustic guitar album! That’s not something I can add”

 Is there anything you feel is left to explore technology-wise?

“Machine learning is developing rapidly and there’s really a lot to still explore there. So I’m still working with it. It may not be the main theme of what I’m doing but each time I’m doing something I’m adding to my toolkit. There are too many brilliant people working on this technology right now. It’s too exciting not to follow the story as it develops.”

So no turning back for you? No stripped-back album on the cards?

“Nobody needs me to record an acoustic guitar album! That’s not something I can add to the conversation. What I can add is to do with deeply investigating technology and how it affects the creative process and aesthetics and the wider music community. I’m also downstream of the research that happens. And I’m always exploring new things and peers.”

Your albums are progressively more melodic. Is that a linear journey?

“Hmm. A lot of the noise stuff, when I was younger, was about flexing in a certain way, but I’ve become more interested in creating entry points for people. That’s not about giving people what they want but I also don’t want to create a barrier between people and my work. Maybe I’ve just mellowed out as I get older. But who knows, maybe I’ll be creating some raging noise again when I’m a granny!”

We’ve talked briefly about the choral ideas in Proto – like many hypnagogic-style records, it visits a few places in music history along the way. What was your touchpoint?

“When I was looking at machine learning and thinking of it as a human coordination technology, it was asking me to think of how music has developed historically in different parts of the world. Similarities and differences. So I was looking at a lot of different music and functions of it over the centuries. So it wasn’t really about one period of time. I guess I was just looking at the function of song in a variety of cultures.

"I guess one is ‘sacred harp’ tradition, referenced directly in the song Frontier. This is a tradition that came out of Britain/Ireland and the American South in rural communities. It’s a particular notation system – called shape notes – that allows people who don’t read music to understand more easily because the shapes communicate the intervals. It’s not really performed for an audience – people stand in a square, singing really full throttle for each other.

"That was something I was interested in, as opposed to the well-rounded European idealised vocal, I was more interested in the folk, nasal voice. Less rounded. So I was tapping into a lot of different traditions, but sacred harp was the most audible. But it was less about one period and more about a shared human archive.

"That’s what I think is most crazy of all about machine learning. As soon as something becomes captured in media it becomes machine legible. So that means our entire archive is legible. It becomes canonised and can create new works from its internal logic.

"Trying to input this shared human history is a great responsibility. It’s something we have to think about as researchers. How we create this shared archive. If we can use a training cannon, for example, training an auto-encoder on Bach’s repertoire so that it can create endless work in that style forever, what does that mean for him or his family or his legacy as a person who was alive at some point? It opens up a Pandora’s box. Super messy.”

What’s next – if any of us can even tell at this point?

“I actually hope that things don’t go back to normal and this is a call for some kind of positive change. But, of course, as soon as we are able to travel again, I hope people find ways to go and support musicians that are sustainable. The climate crisis as well. It’s bringing up lots of grander scale future questions. My partner and I have been thinking about interdependent music instead of independent music, so the name of the podcast is Interdependent. Thinking about how people – in general and in music – depend on each other.” 

Proto is out now on 4AD. Holly Herndon and Mat Dryhurst’s podcast Interdependent can be accessed at: patreon.com/interdependence

Get over 70 FREE plugin instruments and effects…
…with the latest issue of Computer Music magazine