A working vocabulary for synthetic representation. The taxonomy groups these AI moving-image experiments not by tool but by the kind of slop each approach affords. Pick a series, then move through its videos.
slop
Asking where a quality enters could be the whole method.
failure
excess
Identity
Possession
The body stays; the identity is replaced.
6 pieces
The Averaged Phantom
Identity overwhelms the differences between things.
2 pieces
Surface
Boil
The image cannot hold its surface.
4 pieces
The Coating
The image cannot escape its surface.
3 pieces
Presence
The Refusal of Absence
What should be empty is filled.
2 pieces
Overgrowth
What is present multiplies past sense.
1 piece
Category
Feral
The material escapes the label.
1 piece
Second Nature
The label overrides the material.
2 pieces
Event
Still Life
The world refuses to happen.
2 pieces
Perpetual Motion
Nothing that happens changes what can happen next.
4 pieces
Possession
The body stays; the identity is replaced.
6 pieces
The Averaged Phantom
Identity overwhelms the differences between things.
2 pieces
Surface
Boil
The image cannot hold its surface.
4 pieces
The Coating
The image cannot escape its surface.
3 pieces
Presence
The Refusal of Absence
What should be empty is filled.
2 pieces
Overgrowth
What is present multiplies past sense.
1 piece
Category
Feral
The material escapes the label.
1 piece
Second Nature
The label overrides the material.
2 pieces
Event
Still Life
The world refuses to happen.
2 pieces
Perpetual Motion
Nothing that happens changes what can happen next.
4 pieces
Notes on Slop
This is a taxonomy under construction, not a definition settled in advance. The examples approach slop from different directions, and some may turn out not to be slop at all. Those boundary cases matter. Synthetic, unstable, excessive, or uncanny are not necessarily synonyms for slop.
The categories come from making the work. Breaking the videos down into their prompts, source images, constraints, intermediate states, and sequences makes it possible to ask where a particular quality entered the image. Sometimes it seems to belong to the model; sometimes to the procedure I imposed on it; sometimes to the movement between frames. Provenance helps distinguish these cases.
The taxonomy also owes something to photographic typologies such as August Sander's People of the 20th Century and Bernd and Hilla Becher's studies of industrial structures. Their systems make similarities visible without eliminating the awkwardness of individual examples. The tension between a category and the thing asked to occupy it is useful here. I am interested not only in examples that neatly demonstrate a kind of slop, but in the ones that put pressure on the category itself.
Working with generative systems is part of the inquiry. I am interested in their foreignness: the representational habits produced by systems trained on enormous collections of other people's images. Making images with them provides a little purchase on those habits. The point is not simply to identify what the machine gets wrong, but to discover what it considers sufficient, what it insists on adding, and what disappears when a representation succeeds on its terms.
Every kind of slop is the shadow of a loss function.
A loss function is the model's measure of how wrong it is — the number training works to shrink. Reducing some errors while tolerating others teaches the system what matters and what can be thrown away.
The works here look at what comes out the other side: what survives, what disappears, and what the system invents to make up the difference.
Underneath runs one empirical question: what can change without changing what the image is? A face smears and stays Björk; a napkin is remade moment to moment and stays water; a watercolor manner survives with no watercolor behind it. Each case shows which differences the system treats as consequential and which it discards — the loss function made visible, no mathematics required.
Genera are grouped in commitment pairs — a failure pole and an excess pole; family (technical lineage) is recorded but not browsed here. The full chronological body of moving-image work — analog, procedural, and AI — lives in the animation archive.