Creative Fusion

There’s a persistent question surrounding AI-generated imagery:

Did the machine really create anything?

It’s an interesting question, but I suspect not the most interesting one.

For this latest experiment I started with just two images.

One is almost absurdly simple: a handful of black blobs on a white background. The other is a dramatic, adjusted photograph of a tree. Those two images become the only visual ingredients fed into the system.

Everything else is exploration.

By changing the relative strength of each image, adjusting denoise values, varying conditioning and simply choosing different random seeds, an astonishing variety of images begins to emerge. Some feel painterly. Others become surreal. A few hint at landscapes or creatures that were never present in either source image.

The machine isn’t retrieving these from a hidden library. Nor is it 'imagining' them in the way humans usually mean the word.

Instead, it’s navigating an immense space of possibilities constrained by the relationship between two simple starting points.

Watching those images appear reminds me less of painting than of jazz.

A jazz musician doesn’t begin every solo from nothing. They begin with a melody, a chord progression or a rhythm. Creativity lies in discovering unexpected pathways through familiar territory. Every improvisation occupies a different position within the same underlying structure.

This feels remarkably similar.

The two source images become something like a visual chord progression. Each seed explores another route through the same landscape. Some journeys wander only slightly from the originals. Others travel so far that it’s difficult to recognise where they began.

Perhaps that’s why I find these systems fascinating.

Not because they replace artists. Do they?

Not because they make 'better' pictures. Do they really?

But because they reveal just how enormous - and rich - the space between ideas can be.

The accompanying soundtrack follows exactly the same philosophy. It begins by feeding the two source images into MetaSynth’s image synthesiser, converting visual information directly into sound before evolving into the finished piece.

In a sense, both the music and the imagery are asking the same question:

How much creativity exists between two simple ideas?

Perhaps more than we ever imagined.