01 / The starting point

Collecting footage

The project began with Sara Fox’s dance video, a little over 54 seconds long. Mariano Zamora’s 38-second clip later added a second voice. The aim was to carry its pauses, turns and bursts of energy into an artwork. Keeping the original timing mattered: a quick gesture should still feel quick, and a held pose should have room to breathe.

02 / Reading movement

Turn frames into a movement score.

Python and OpenCV helped read the video. MediaPipe Pose Landmarker estimated 33 body points in each frame, including the shoulders, hands, hips and feet. Their changing positions gave us direction, speed and acceleration. Small jitters were smoothed, and short gaps were filled carefully. The tracking is an estimate, so the overlay helped reveal where it needed checking.

The results were saved as timestamped JSON: a file of movement data the browser can follow. The video remains the clock, so the field stays in step with the dance.

03 / Finding the material

From a magnet to something she could move through.

The first study used a simple magnetic field. Its curved lines were beautiful, but one central magnet felt too still. Adding the hands and feet brought more of the choreography into the field. Body contact made the next difference: particles could be pushed aside by a kick, flow around an arm and return into the space left behind.

Dozens of iterations followed. Graphite shading brought out dense ridges while keeping individual grains visible. Blue currents then made active areas easier to read, using the same moving particles.

04 / Light within the field

Let the densest places glow.

The rods stayed small and sharply defined as the surface changed from white filings to graphite and then blue light. Brightness gathers where particles are actually close together. A tight glow softens the brightest cores; faint resting rods give the surrounding darkness depth. These are changes to the material’s appearance, built on the same movement.