Skip to main content
Baryon feeds the incoming waveform into a modeled acoustic cavity. The renderer reveals a thin, luminous structure derived from the resulting resonances.

From a full mix to a visible structure

  1. Baryon captures consecutive audio samples, preserving mono or stereo input.
  2. Signal preparation removes DC and applies the configured weak water-drive correction.
  3. The prepared waveform drives a fixed bank of damped resonances. Each resonance retains its pressure and motion state as the music changes.
  4. The engine combines those responses, including their interference, and measures pressure and motion over a short time window.
  5. A spatially smoothed balance of those measurements defines a surface. Persistent geometry, spectral color, and light integration give it its visible form.
A DJ mix can contain kicks, vocals, synths, cymbals, reverb, and noise at once. They enter together through the waveform. Each resonance responds to the waveform rather than being assigned to one musical note. The finite mode bank limits the spatial detail of the modeled response; it is not a limit on how many input frequencies can contribute. The modeled cavity responds more strongly at some frequencies than others. Receiving the full waveform does not mean reproducing every sound with equal spatial accuracy.

What is the luminous surface?

The surface follows U = 0, where the weighted, time-averaged pressure and motion contributions balance. The field is spatially smoothed before the surface is observed. This is not necessarily zero sound pressure, a force balance, or a place where real particles would become trapped. Pattern Persistence controls how the observed geometry evolves over represented audio time. When audio ends normally, stored response can ring down. Pausing holds the acoustic timeline; stopping or replacing the source resets the relevant history. Brightness follows raw response energy through the material’s fixed optical mapping. In Spectral mode, color comes from the response spectrum through an artistic octave-repeating palette. Sound is not physically converted into visible-light frequencies.

How physical is the water model?

The virtual apparatus is a rectangular water cavity with a default acoustic side length of 6.25 m, a declared sound speed, and passive damping. Neumann and Dirichlet settings change the wall conditions. The spherical view clips the rectangular field; it does not turn the apparatus into a spherical resonator. A declared incident-pressure reference controls weak waveform steepening and harmonic generation before the linear modal response. It is a calibration parameter of the acoustic model, not a measurement of your speakers or room. Baryon combines established acoustics with a chosen observation and artistic material; it does not simulate fluid flow, cavitation, particle transport, or the full motion of a laboratory water surface. For stereo, the engine responds to each channel separately and averages their pressure/motion measurements. This preserves both channels without canceling opposite-polarity material through an initial mono downmix. It models two virtual experiments, not the placement of two loudspeakers in a room.

Why settings matter

Most controls are either analysis controls, field controls, render controls, or camera controls. The carrier core’s world-space FWHM is fixed by Baryon’s internal reference-apparatus profile. It is not a user control or an automation target.

Camera versus rotation

This distinction matters when you are trying to capture a specific expression:
  • Camera orbit changes where you view the field from.
  • Rotation Mode changes whether the rendered cymatic volume itself rotates.
  • Manual Spin Speed only applies when Rotation Mode is Manual.
  • Audio Rotation Amount only applies when Rotation Mode is Audio.
  • Lock Camera prevents accidental orbit dragging. It does not freeze the cymatic field.

Choosing an audio input

Use a direct line or loopback feed for a DJ mixer or another music app, a file for repeatable playback, or a microphone for room sound or a live source. Live-input classification describes the capture setup. The retained Voice/Vocal and Ambient intent labels do not select a single pitch or change the number of frequencies admitted by the broadband solver. For the most faithful result:
  • Use a clean recording or a direct feed when possible.
  • Keep echo cancellation, noise suppression, and auto gain off for line feeds or clean files; turn them on only when a noisy room makes capture difficult.
  • Start with lower Material Density and Caustic Strength, then raise them until the field reads clearly without flattening into a bright sheet.
  • Save a preset per source so you can return to a look instantly.

What coordinates would mean

Because Baryon renders a continuous field, an exportable “coordinate” set is not the same thing as a point cloud. A future geometry export would need to choose a moment in time, sample the field, extract a surface or volume, and convert that result into a mesh format that design and slicing tools understand. The current public app does not export raw coordinates, point clouds, STL, OBJ, GLTF, or other 3D-printable geometry. See Creative workflows for the current capture path and the shape of a future export workflow.