It’s hard to have a conversation about creative AI without getting existential, and no one knows that better than Portrait XO. Since the 2022 release of her AI-powered debut album, Wire, the Berlin-based creator has watched as the conversation around generative AI exploded worldwide – bringing with it a deluge of questions not only about her creative process, but about what it all might mean for the future of creativity, and the human condition.

“It's intense,” she concedes. “But I’m not surprised – AI is a rabbit hole and it’s impossible not to get drawn into the deep philosophical questions that surround it. Everyone's hungry for perspectives, and it’s the role of artists to show the possibilities and limitations.” 

Far from shrinking from such questions, the transdisciplinary musician draws energy from the debate. She’s appeared at the UN AI for Good Summit, the recent Reeperbahn Music Festival in Hamburg, and myriad other conferences to give her perspective on how creative industries can chart a course through what she expects to be a decade of disruption.

“No other technology I’ve worked with provokes these deep questions about policy, legality, philosophy, and politics,” she reflects. “What makes the topic of AI and creativity so different is that people can see we’re at the very start of the next era. Right now, we’re figuring out policy and regulations that will impact the legacy of this technology for the next decade or more.”

It’s for this reason that Portrait XO emphasises the importance of independent artists in the conversation – to restrict the debate to the biggest voices risks repeating the mistakes of the past.

“My hope is that we don't create another problem like Spotify,” Portrait XO says adamantly, noting how the streaming platform ultimately reinforced the industry’s historic power imbalance over its artists. “The on-the-ground decisions that were made were a mistake that left a legacy because once that cultural impact has happened it’s really hard to shift. So, how do we prevent that from being repeated?”

“Labels, meanwhile, are all working on stuff on the back-end: they have master rights to a huge back catalogue of music, and their own in-house data scientists,” she says, noting that this gives them enormous power to create new music out of thin air, with ‘factory-produced pop songs’ already flooding streaming platforms. “Being an independent musician who isn’t signed to a label and has full ownership of their material is an incredible positive right now.”

Figuring out what society, as a collective, wants and will accept, is key. “I'm in the process of pulling in different industry experts to work on creating an open framework, a navigational place, for an evolving document that addresses some of these legal implications and concerns. We need to allow both experts and the public to transparently share opinions; we can’t rely on a small set of humans to make these important decisions, we really have to work together.” 

Of course, none of this was on her mind back in 2019 when she initially began experimenting with AI alongside visionary music-makers, Zack Zukowski and CJ Carr, of DADABOTS.

“They started nudging me about this AI research they had been working on,” she recalls. “2015 to 2018 were really weird years for me in terms of music; I was writing so much material but I was unhappy with the majority of it. I kept finding myself in styles that felt confining, I was desperate to break free from those forms.”

Using a customised version of the open source SampleRNN model, the trio began to break the creative deadlock. “The AI model took a hammer and cracked all my patterns and habits,” she says of the experience. “Everything broke into a billion pieces and then it gave me back these amazing fragments.”

What makes the SampleRNN model so special is its ability to combine melodies, lyrics, and vocal techniques in its generative output. Trained on Portrait XO’s voice, the model returned some 10 hours of audio, broken into 60 second clips.

“About 10% of that audio had cohesive lyrics and melodies, and the rest were glitchy, weird artefacts.” Even within that hour of usable audio, the generated output was full of half-finished sentences and unusual vocalisations. While some might have found the process frustrating, for Portrait XO, those snatches of song re-inspired her own musical imagination, becoming what she fondly describes as a “creative sparring partner”. 

“Every time I heard a lyric, a melody, or phrase that I’d never sung before – something that I would never normally sing – it would inspire my songwriting process to work in a different way. It became this call and response.”

“AI became my creative sparring partner.”

— Portrait XO

The resulting work would go on to form the basis of Wire, Portrait XO’s first full-length album release. Drawing oblique influences from artists like Bjork and Portishead, Wire pulses slowly across its ten tracks. Ample space is left for Portrait XO and her AI counterpart to sing in harmony as scattered percussion and synth tones fill out a darkly polished production. 

Since the album’s initial release in December 2022, Portrait XO has followed up with music videos for the tracks VESSEL, and AIV1, each using generative AI and visuals which react dynamically to the features of the music, in order to add a distinct new dimension to the music.

Released today and available to watch below for the very first time is Sitting on A Wire, Portrait XO’s third release from the album and easily her most ambitious audiovisual project to date. 

Combining the opening two tracks from Wire into one song, the video blends live footage of the natural world with a panoply of visual effects, post processing, and text-to-image AI. As the music shifts seamlessly between Portrait XO’s human voice and her AI counterpart, so too does Sitting on A Wire depart the physical realm into a place of restless shapes, impermanent terrain, and lines of melting gold. 

“I've always loved films and animations that give the viewer a sense of being between reality and virtual reality,” she says, noting the music video’s ethereal aesthetic. “I wanted to tell the story of human-machine creation, and to show the hybridity of that journey.”

The video has been an almost two year journey, a process that “reached points of insanity” at times, she admits. In truth, the creative friction of working with AI – and the level of effort and determination required to achieve satisfying results – is a theme that comes up repeatedly in our conversation.

“We want to be challenged.”

— Portrait XO

She likens the experience to that of learning, or of mastering an instrument; the learning curve may be steep, but the eventual payoff is a new medium of self-expression.

Ultimately, the technical hurdles will be ironed out – but perhaps part of the value in any craft is the artistic struggle.

“AI tools will make music creation more accessible to the general public, and that's awesome, but on the other hand I think it's actually good to not have the entire creative process be super easy,” she says. “We want to feel some kind of friction, we want to be challenged: I don’t think we would be happy if everything was too easy.”

“Everything broke into a billion pieces and AI gave me back these amazing fragments.”

— Portrait XO