SL-52 / The Rhythm Set

Humming in Canon

Two people alternate phrases, then join. The audio leads; the video obeys.

The brief

Two people alternating hummed phrases and then joining together, with the piece designed audio-first.

Why it's hard

Video models generate sound as an afterthought decorating pictures. Here the sound is the structure, and the bodies have to be answerable to it.

The craft key

Audio-first design: write the canon, then make the video obey it.

Skill Tests

04 / 28

Animal anatomy Human anatomy Biomechanics Physics & mass Water simulation Crowd dynamics Multi-shot continuity Face consistency Camera movement Unbroken take Motivated lighting Cutting rhythm Hands & fingers Hair & fabric Reflection & refraction Scale & depth Impossible geometry Period accuracy Music sync Aging & time Natural asynchrony Eye-lines & gaze Particle behaviour Motivated light source Consequence & persistence Time manipulation Graphic-layer discipline Colour as storytelling

Every challenge is scored against the same vocabulary, so you can compare them. Lit cells are what this one tests.

Starter prompt

Built with the CEMI MediaMax system for Google Veo 3.1. Prompts stay in English — that is the language these models take instructions in. Treat it as a starting point and start varying it.

humming-in-canon

Two people shelling beans on a porch hum a simple four-bar phrase in canon at about 66 BPM. One starts alone. The second enters exactly two bars later on the same phrase, a third lower, so the lines overlap for two bars. Neither is quite in tune; one runs slightly ahead and pulls back. On the final phrase they arrive together, then stop unevenly — one holds the last note about a second longer. Build the picture to the audio: hands keep working through the whole take and neither looks at the other while humming. Shot in Apichatpong Weerasethakul's patient register, lit like Agnes Godard with daylight from one side only. 50mm vintage prime at f/2.8 on a tripod, with one slow reframe when a bowl slides toward the table edge. DIRTY FOREGROUND: screen-door mesh across the left third. Materials: split cane chair, chipped enamel bowl, dry bean pods.

Derived from the CEMI MediaMax production canon — cemi.media

Style reference: Apichatpong Weerasethakul, lit like Agnes Godard

Why it's built this way

Canon is a timing relationship, so the entry point is given in bars rather than described as harmony, and the intonation drift keeps it off the quantised grid. Locking the hands to continuous work means the audio can lead without the bodies turning into a performance for the lens. The single tripod reframe is the only operator gesture, motivated by a bowl about to fall.

Watch for

Both humming in perfect unison from the first bar, faces turned toward each other and toward camera, mouths moving with no relationship to the audio track.

Ran a challenge? Show us what you got.

Share your attempt — the triumphs and the glorious failures. The most instructive entries get featured on the board.