A sound,
unfolded.

The 7.4-second phrase is synthesized deterministically from an explicit musical score with glass-and-mallet-like additive tones. It is an original interface study, not an excerpt from Amaç Erdem’s compositional catalogue. No downloaded recordings or sample libraries are used.

MONO PCM · 24,000 Hz · 16 BIT · 7.4 s

The spectrum is computed from the exact exported PCM samples. A 2,048-sample Hann window moves in 1,024-sample steps. The display uses 64 logarithmically spaced bands from 80 to 8,000 Hz and 174 time frames.

Each band uses its largest available Fourier-bin amplitude, corrected for the Hann window’s coherent gain. Very narrow low-frequency bands can share a bin. Values are displayed over a fixed −60 to 0 dBFS range; the reference is a full-scale sinusoidal peak amplitude. There is no separate normalization for each frame.

X: TIME · Y: LOG FREQUENCY
RELIEF & COLOUR: SPECTRAL MAGNITUDE
FOLDING & LIGHT: ARTISTIC TRANSFORMATION

Decorative ribbons interpolate between measured bands. They do not increase the analysis resolution. The sculpture is neither the literal shape of sound nor a neural measurement.

A single ribbon mesh travels through seven states: a folded surface, an open spectrum, converging contexts, possible continuations, an interwoven structure, a score-like surface and a returning ring. The mesh retains the same vertices and connectivity across the journey.

Only the open spectrum uses literal time and frequency coordinates. The other forms are authored visual metaphors; they are not model outputs, brain anatomy or measured cognitive states. The light travelling over the material follows the playback clock, not a physiological measurement.

The three continuation controls highlight corresponding branches. Their dimensions do not encode confidence or likelihood. A still-view control removes drifting, pointer motion and the travelling playback light.

The two three-second lead-ins differ in pitch content, voicing and implied tonal centre. Each is followed by an 80-millisecond gap and the exact same 1.4-second A4/E5 target file, scheduled on one audio clock.

The activity invites a personal comparison. It is not a controlled empirical study, and a difference in experience is not assumed. Your response stays in the page’s temporary state and is not sent to a server.

Return, Suspend and Open are three composed synthetic examples. Their curves are illustrations, and their order and visual weight do not encode model probabilities.

The study is a starting point for asking how a continuation changes a phrase. A model-based comparison would need its own implementation, inputs and evaluation.