Audio-reactive texture
The audio output drives a visualizer layer that captures bass / mid / treble energy, RMS, beat detection, and recent-note history. Compositions can shape themselves around their own loudness profile.
The visualizer is not just a screensaver. It is a sampling layer that watches the audio output in real time. FFT analysis, RMS, spectral centroid, and beat detection are all available to the rendering side, all updated frame by frame. The composition produces audio; the audio drives the visual; the visual responds to the music.
What the visualizer captures
- Bass / mid / treble energy: three FFT bands smoothed with a 0.85 decay.
- RMS: overall amplitude; smoothed.
- Spectral centroid: perceptual brightness; useful for distinguishing percussion-heavy from melody-heavy passages.
- Beat detection: onset events with intensity.
- Recent note history: every note that has played in the last few seconds with timestamps and pitch.
Composing with the loop
Today the audio-reactive layer drives visuals only: particle fields that pulse with the bass, brightness shifts that track the centroid, beat-aligned ripples on the visualizer surface. Signals through the graph are server-authoritative and are not influenced by the audio analyzer. Treat audio reactivity as a listening surface: it makes the mix visible without letting the visual layer rewrite shared room state.