ANIME-GAN [2018]

ANIME-GAN [2018, ai/crypto]

⋆ Format: .mp4

⋆ Dimensions: 5120px x 5120px
⋆ See/purchase:
1/1 [Edition on one]

Reflecting on the trajectory of AI-generated imagery, from its early experimental roots to its current state of mass-production, these days I find myself more than ever returning to the ethos that first drew me to the medium and excited me endlessly to learn and understand it. In 2018, training a GAN on anime films [including those from Studio Ghibli], and I remember to me at least felt akin to operating a personal printing press [which at the time one I was running Wounded Wolf Press surely must have played a role making digital editions] a process defined by slowness, a bit unpredictability, and an intimate negotiation between artist and machine. Each output was a singular, emergent creation, an artefact of collaboration rather than automation which you have to hone to maths of the training process and model.

When I first worked with GANs, I was captivated by how machine learning could take fragments of this aesthetic and reassemble them into something new, often hallucinatory, half-formed, yet deeply evocative. It felt like unearthing forgotten frames from a film that never existed, an archaeology of latent space that resonated with the way animation itself often grapples with memory, nostalgia, and the limits of perception. In this case, despite how GANs typically function, often trained and generated within a single class, here the results are a mush of anime aesthetic and indistinct imagery which has the soul of anime without any semantic content, yet it was precisely this quality that made it compelling. Unlike most artists and researchers working with GANs at the time, who focused on training models to produce coherent, dataset-specific outputs, I approached it as a means of abstraction, using the technology to create something closer to expressionist painting rather than mere replication.

So, I feel this work stands in stark contrast to the current AI landscape in many different ways, where tools such as OpenAI’s image models have catalysed a flood of Ghibli-fied images, homogenising a once-singular aesthetic into an infinitely reproducible template. As these images overwhelm platforms, OpenAI itself buckles under the sheer computational demand, mirroring the shift from small-press zines to an endless feed of frictionless, surveillance-capitalist content. AI, once a conduit for artistic interrogation, is increasingly deployed as a means of mass visual production, its output less about process and discovery, more about immediate, algorithmically-driven gratification. Here, I almost can’t believe myself where I become a critique of any sort of AI driven enhancement or social movement, I’d celebrate this blindly a few years back, but as noted around my other collection RXALITI too, I feel the only poetic and artful happening around this Ghibli AI art was the fact that the demand was so much it melted the GPUs at the OpenAI. This was precisely where RXALITI was always headed: a prophetic vision of the web’s agentic decay, where neural networks churn out spectral remnants of the past, hollowed-out derivatives masquerading as artistic progression. RXALITI wasn’t about AI “enhancing” creativity, it was about AI parasitizing itself, collapsing into self-referential sludge. The Ghibli trend exposes this in real-time: art reduced to an endlessly replicating style filter, a sentimental ghost of what was once an artistic movement rooted in human touch and philosophy.

RXALITI’s dead forest theory of the internet saw this coming, the space is overrun with synthetic growth, where each piece of AI-generated art is another leaf in a decaying ecosystem, drawing from the same soil of overtrained datasets and recycled cultural memory. We’re watching the collapse happen in front of us, yet we continue to engage, boosting the machine that feeds on the remnants of real creative movements.

This NFT, minted now, I hope asserts itself against that tide. To me, it marks a return to AI as a space for intention, depth, and critical engagement rather than a tool for automated pastiche and serves as a reminder even for myself. By choosing to mint and release this work at this moment, I personally seek to highlight the urgency of this conversation: What does it mean for artists to work with AI as a material rather than a generator of aestheticised excess? How can collectors engage with AI art in ways that prioritise artistic integrity over ephemeral trend cycles?

And perhaps most importantly, as I sit by my window listening to bird songs, I want to think this is a call to embrace slowness, to recognize AI not as an engine of infinite output, but as a medium shaped by its own tensions, failures, and idiosyncrasies. It is a space where art can still be made with care. We should celebrate its possibilities, much like the ‘sound card revolution’ that brought music production into bedrooms, allowing artists to create entire albums independently, except this time, AI offers that potential for cinema, for visual storytelling, for entirely new forms of creativity.

But this must be approached with clarity, with an awareness of what fosters true connection and where human agency risks dissolving into the noise. Art, at its most potent, exists on the edge of that dissolution, where meaning and chaos blur, and signal becomes indistinguishable from static. Like RXALITI always sought to explore this threshold, introducing deliberate noise into the AI model to test where the boundary between recognition and entropy lies, I wish this work will make a case by showing the spirit of early exploration and AI as an expressive tool.

I never wanted to fully agree with Fisher’s hauntological diagnosis, the idea that we are trapped in the ghostly loops of lost futures, endlessly replaying fragments of what could have been. But maybe this is the moment I start to see the edge of it and almost a cyber-hauntological method of curation and way of thinking about art. The algorithmic churn, the infinite recombinations, the way AI-generated art mirrors cultural memory in a way that feels both expansive and recursive, perhaps RXALITI and these editions of one works, in its own way, are an attempt to grasp at something beyond that loop. This piece, I hope, will act as a signal, a moment of clarity within the vast, accelerating churn.

Thank you <3

Atay [31.03.2025, the year of soul-rot]