Total Pixel Space: Finite Pixels, Infinite Debate
RunwayML have done something both inevitable and fascinating: launched an AI Film Festival. For a few years now, people have been speculating whether AI could meaningfully create not just images but whole moving pictures, stories, and cinema. We’ve had experiments, memes, uncanny valley horrors, and a few moments of brilliance. But this is different. This is an organised, curated space with entries, judges, and winners.
This year’s winning film was called Total Pixel Space, a short, mesmeric exploration of what it means to be “unique” in a digital medium where every image is simply a finite arrangement of coloured dots.
And that’s the central idea I thought was worth looking at. Because in an age where “AI art” sparks furious debate about originality, theft, plagiarism, and the very definition of creativity, Total Pixel Space asks us to step back and consider the fundamentals: if everything is ultimately finite, how do we define originality at all?
Every image on a screen is, at the lowest level, a grid of pixels. A limited set of tiny squares, each displaying a colour from a limited palette. Combine them together and you get a photograph, a painting, a screenshot, a meme. Change them again and you get another.
The possible combinations are astronomical. But still, finite.
This is the paradox. Rearrange the pixels enough times and you will eventually land on every possible image, from the Mona Lisa to your exact face staring back at you, to an image that has never been seen before. The space of possibilities is vast, so vast that it feels infinite, but mathematically it isn’t.
When you extend this into moving images the number multiplies exponentially. One frame may already contain trillions of possible arrangements. Now multiply that by 24 frames per second, for minutes or hours. It’s so big it’s incomprehensible. And yet, still not infinite.
That’s the real trick of Total Pixel Space. It shows us that uniqueness in the digital age is always framed against this vast but bounded canvas.
But it isn’t new. Music has been wrestling with it for many years.
Western music is built around twelve notes. Add rhythm, harmony, dynamics, tempo, and you get millions of songs, more than enough to keep us entertained across generations. But at its core, the building blocks are few.
Which is why plagiarism in music keeps reappearing. There are only so many chord progressions, only so many ways to string together melodies that sound pleasing to the human ear. Some overlaps are genuine theft. Others are accidental. And some are fabricated disputes that say more about lawyers and publishing rights than they do about artistry.
The legal battles faced by Ed Sheeran made this tension obvious. Sheeran himself pointed out that coincidence is inevitable when the raw materials are so limited. The question isn’t whether two songs overlap, of course they will. The question is whether the overlap is meaningful.
Art has always existed within these constraints. So why does AI-generated art stir such visceral reactions?
Partly it’s because of the speed. Human creativity, even at its most prolific, is slow. Writing a novel, painting a canvas, composing an album - these take time, skill, and persistence. An AI can generate thousands of images in the time it takes to make a coffee. That sheer acceleration feels threatening.
Partly it’s about authorship. With a painting or a song, the chain of creation feels clear: a human took the time to learn a craft, to express themselves, to put something of their inner world into the outer world. With AI, the waters muddy. The artist is the prompter, not the brush-holder. Does that count? Is it lesser? Or is it just different?
And partly, it’s because we’re still figuring out value. If anyone can generate something polished-looking in seconds, then scarcity collapses. The flood of so-called “AI slop” [lazily generated, poorly prompted outputs] drowns the signal in noise. That cheapens the perception of the whole, even when genuinely powerful, moving work is produced.
But this, again, is not new. There have always been amateurs.
Plenty of people buy guitars and never learn more than a few clumsy chords [speaking from my own experience here!]. Plenty of people open Photoshop and produce something ugly. Plenty of people write songs, paint, sculpt, or make films that never get beyond mediocrity.
We don’t judge the whole medium by its weakest practitioners. Nobody says “painting is terrible” because a thousand people make bad paintings. Nobody says “music is dead” because of every out-of-tune pub band.
What matters is the difference between the artisan and the amateur. Skill, craft, intent, persistence, refinement. A guitarist with no training might strum noise. A trained musician might take the same three chords and make something transcendent.
AI is no different. A poorly written prompt produces junk. A carefully constructed prompt, iterated with intent, can unlock extraordinary results. The artistry isn’t in the wielding of a brush. It’s in learning the nuances of coaxing, guiding, and manipulating the machine to channel imagination.
It’s not “cheating.” It’s simply a new tool.
Another reason AI art unsettles us is that it exposes the replication at the heart of all creativity.
Humans are magpies. We remix what we’ve seen, heard, read, felt. Every novelist has influences. Every filmmaker borrows shots, framing, tropes. Every musician carries echoes of others. We like to tell stories about originality as if ideas spring fully formed from the void, but they don’t. They come from drawing on experiences and existing sources.
AI makes that process explicit. A model is trained on vast datasets of existing work. Its outputs are mathematical combinations of patterns it has seen before. Critics call this “plagiarism.” But isn’t that just a mirror of human learning?
Perhaps the difference is scale. A human can only absorb so many influences in a lifetime. A model can absorb millions. That breadth feels alien, and maybe unfair. But the principle is the same. Everything new is stitched together from what already exists.
Which brings us back to Total Pixel Space.
If every possible image can, in theory, be generated eventually, then originality isn’t really the point. Uniqueness is an illusion at the level of raw materials. What matters is what resonates.
A song isn’t meaningful because it uses a chord progression no one has ever touched before. It’s meaningful because of how it makes you feel. A film isn’t powerful because its pixels form a mathematically novel arrangement. It’s powerful because of what the story awakens in you.
AI or human, prompt or brushstroke, it’s the same. What we respond to is the spark, the connection, the human experience embedded (however indirectly) in the work.
AI art isn’t going away. Just as photography didn’t kill painting, just as synthesisers didn’t kill acoustic music, just as word processors didn’t kill writing. Every new tool disrupts. Every new tool causes panic. Every new tool forces us to re-evaluate what matters.
Yes, there will be oceans of “AI slop.” Yes, there will be ethical battles about training data, credit, and ownership. Yes, we will need new frameworks for value, attribution, and curation.
But there will also be works like Total Pixel Space - thoughtful, provocative pieces that use the tool not just to dazzle, but to ask meaningful questions about creativity itself.
Maybe uniqueness is overrated. Maybe the pursuit of novelty for its own sake is less important than the pursuit of resonance, clarity, and meaning.
Because in a world of finite pixels and finite notes, what lasts isn’t what’s “new.” It’s what matters.
Watch the movie and let me know what you think : https://www.youtube.com/watch?v=OiZ7XTbf2cg
