How it works

How MusicDNA works

Four steps from sound to signature.

Choose a track
Pick any audio file from your music folder.
Decode the audio
We decode the file with the Web Audio API in your browser.
Extract real features
Rhythm, energy, spectrum, harmony, and structure are extracted from the signal.
Generate visual DNA
Deterministic visualizations are rendered from the real analysis.

Visualizations are deterministic. The same file always produces the same DNA.

What MusicDNA reads

Every visual is rooted in measured signal features.

Downsampled peaks and RMS envelope across the track.

Onset strength, beat times, and BPM estimate via autocorrelation.

FFT frames, spectral centroid, flux, and low/mid/high balance.

Chroma vector and key estimate via Krumhansl-Schmuckler profiles.

Section boundaries from spectral and energy novelty.

Loudness curve, dynamic range, loudest and quietest moments.

Not every feature is available for every track. Uncertain values are clearly labelled as estimates.

See the seven visual modes these features drive. Explore the examples