Composition Films: Ekphrastic Coding

Composition films draw on tagged archives and a rule-based grammar for how images and sounds may meet in time. Rules read hand-selected metadata and compose in real time—each screening one path, not a stored master cut. What the films hold in common is the material and the grammar; what varies is the composition each run produces.

Xri Xro, my first composition film, serves as a simple example. I have tagged the first (xri — “roll in”) and last (xro — “roll out”) shot of every roll in my archive^[1]. The single rule of Xri Xro is to oscillate between these two pools; the effect is a flare in from white to image, cut to image, flare out to white.

In my latest film, Of Rivers, shots are categorized subjectively by two perceptual axes—proximity to water (w0–w5) and the movement of the water (m0–m5)—and sounds by sonic intensity (ls0–ls5). Two examples: a shot of slow-moving water from far away is categorized as w1 m1; a waterfall up close is w5 m5.

Together, proximity and movement imply a third, image-derived value—intensity—averaged from the two axes; that value selects sounds from the corresponding ls pool. The overlap of these two axes forms a grid of 36 possible water representations.

The core editing schema crosses the grid of states—e.g. from w0 m5 toward w5 m0, or from w5 m5 toward w5 m0—one quadrant step at a time. Which direction the images go is weighted by probability and history. That crossing of the grid is called a movement. At a movement’s end, the path may reverse or jump to another corner.

How the images and sounds unfold in a movement is determined by the modality and mode. A modality fixes how picture and sound relate in the cut; a mode is one inflection of that relationship. One modality is called “layered”: images step through the grid, black leader optional between them; sound either cuts to silence on the leader or cascades over black into the next image—this is determined by mode. Which modality/mode guides the next movement is determined by chance and what preceded it.

To develop the rules for Of Rivers, I reverse-engineered my editing instincts. For example, I wanted a sequence where the images move toward a waterfall as the sound’s intensity decreases–this is coded as a movement from w0 m0 to w5 m5, with sounds moving from ls5 to ls0. Composition films are an ekphrastic encounter. The code is ekphrastic writing—rules answering the footage in language; the screening is reverse ekphrasis—that language answered in time, stochastically.

Chance belongs to the subject, not only the system. In Xri Xro, a roll-in meets a roll-out drawn from eighteen years of footage—de-chronologizing the archive into something closer to memory’s aleatory recall than to chronology. For Of Rivers, I returned to the same stretches of river many times over; the footage proves what the proverb only claims—different speed, different nearness, different light each time. Chance in the composition keeps the edit from fixing what the water will not hold still.

1. 1.As of 2026-08-18, the archive contains 26,682 16mm images spanning 18 years of filming. ↩