A VISUAL SEQUENCE BY WAAVE
DRIFT
The song doesn't change. The film does. Every play is a different edit, made in real time when you press play. Only this once.
The river
There is a Greek line, attributed to Heraclitus. No one steps in the same river twice. Not because the river has changed. Because the person stepping has, and so has the water.
DRIFT is built around that idea.
Each time you press play, a new edit is built on the spot. It draws from a pool of monochrome scenes: ocean, figures in mist, hands at a piano, dolphins, sparks. The song is DRIFT, produced by WAAVE with vocals from Olugbenga Adelakan and incredible drums by Victoria Smith. It plays from start to finish without cuts. The video over it is never quite the one you watched before.
The work, structurally
The song is 4 minutes 24 seconds. The footage pool runs 6 minutes 47 seconds. The editor divides the song into 20 sections at moments where the music turns. Those boundaries come from spectral novelty in the mel-spectrogram, snapped to a phrase grid of 4 bars at 161.5 BPM, about 6 seconds each.
Each section is tagged with a mood. The footage is sliced into thirty scenes, each tagged with the same vocabulary:
For each section, the editor picks a scene at random from the matching pool. The only rule: a scene cannot appear twice in a row inside the same mood. The peak section, at 2:47 into the song, holds the moment of greatest energy in the spectral analysis. It always lands on the trio finale. Everything else moves.
Two video elements alternate. While one is on screen, the other preloads the next clip. They cross-fade for about 450 milliseconds at every section boundary, so the seek is invisible. The audio is the master clock.
The numbers
For a pool of n scenes drawn k times under the no-immediate-repeat rule, the number of arrangements is n × (n − 1)k − 1. Multiplied across all moods:
That comes to about 82.6 trillion. At one play per minute, around the clock, exhausting the structural space would take about 157 million years.
This counts only the structural arrangement. Many scenes are longer than the song section they fall into, and the editor picks a random in-point inside the slack. That part is continuous, so the practical space is even larger.
Each play is identified by a five-character signature in the bottom corner. Write it down if you fall in love with one. If we add scenes to the pool later, that exact edit may no longer exist. Most of them survive every change, like dialects of a living language.
The choice
This was a deliberate choice.
Most films want to be seen the same way every time. They are released. The implicit contract is: this is the canonical version, this is what you saw, this is what we made.
DRIFT works differently.
What you see right now is one passage through something larger. The work is the system that produces it. Every play makes a different passage.
There is no canonical DRIFT. What's fixed: the song, the system, the trio finale at the peak. Everything else is open.
This is true of any film if you think about it. You have never watched the same one twice. Your second viewing, even of an identical file, lands in a different version of you. The shifting edit makes that obvious instead of hiding it.
It changes how watching works. With a fixed film, you compare what you see to a reference version. With DRIFT there is no reference. What you see is the film. Each play is the only one.
The makers
Find more at @waave.orchestra.