The recording is deliberately artless: no edit, no camera, just the page performing for a visitor. What it documents is the part a still can't — the social choreography. When one head speaks, the other eight turn, and the turn is real 3D: each face's features slide across its skull with correct foreshortening, drawn in strokes that never stop boiling.
Both dubs were generated by Kimi K3 along with everything else — the drawings, the code, the music bed, the per-character voices. The 中文 cut is the default because that's the piece's own default; the toggle above the player mirrors the footer control on the live piece, where you can click the heads yourself.
Speech and music throughout. A generated lo-fi music bed runs under the whole film, and one head at a time speaks a short self-introduction aloud — Mandarin in this cut, with an English dub selectable above the player. Each spoken line also appears in a hand-drawn speech bubble, primary language on top, translation beneath.
Fifty-five seconds, square frame, screen-captured at 60fps from the live canvas. Nine doodle heads sit in a three-by-three grid on ivory paper under the serif title If we all had heads. The camera never moves; the performance is the page's own. Heads sway gently and blink in stagger; pencil lines boil. One by one, a head steps slightly forward and introduces itself — Kimi with its sparkle antenna, Claude wide-eyed saying it is listening, DeepSeek waking mid-thought, Mistral going where the wind goes — lips moving to the waveform while the other eight turn to look at the speaker. Between introductions the sheet settles back to its idle sway. The film ends with the cast at rest.