How sound becomes motion: FFT, beats and response
15 posts in Audio & Reactivity
The Waveform engine's area mode closes its trace back along the baseline, fills it translucently, and then strokes the outline on top. Here is why the order matters and what the depth control does when there is only one row.
Piano/Step Grid rounds every column to a whole number of cells and paints the dividers in the background colour, which is why Background is one of the few controls this engine opens up to layers.
Linear Equalizer sits on the frame edge on purpose, rounds only the two corners at the top of each bar, and runs a gloss across the field that brightens without changing a single bar's height.
Frequency Ridge strokes its front row at more than double the width of every row behind it, fills only the first six, and is the one mode in its engine that reads the noise control.
Line Spectrum draws its samples at evenly spaced x positions and still gives you a logarithmic frequency axis. The curve is in the analyser, not in the renderer, and that distinction is worth understanding.
Oscilloscope/Lissajous needs two channels to draw a figure, and a delayed copy of the first one will not do. Here is the filter that produces a genuine quadrature partner, and why the figure used to collapse to a line.
Center Mirror grows bars in two directions from the middle of the frame, which halves the room each one has. Here is the budget that fixed it, and the gloss that brightens bars without lying about their height.
Mirrored Area reflects a solid spectrum shape across its baseline. Doing that on a canvas is not a transform and a second stroke, and the mode is named after the bug that taught us so.
Spectrogram Waterfall draws rows of spectral history receding into the frame. Speed does not change how fast the buffer fills; it changes how far apart in time the drawn rows are.
The other half of how audio visualizers work: requestAnimationFrame, reading AnalyserNode, and drawing bars, particles and tunnels on an HTML Canvas every frame.
What a Fast Fourier Transform actually measures in a song, how AnalyserNode bins become bass/mid/treble, and why those numbers drive every audio-reactive canvas.
The Audio panel is a signal chain, not a pile of sliders. Here is what each control does, the order they apply in, and how to fix a visual that pulses instead of reacting.
Where the free preview loops come from, what music is safe to publish in your own video, and what copyright actually means on YouTube and TikTok.
How Novus turns BPM into discrete visual hits, why silence stays still, and how to tune Beat Response, Pulse Intensity and Shake Amount for your genre.
A plain-English look at what turns a song into moving visuals: the FFT, frequency bands, amplitude and beat detection that drive every audio-reactive scene.