Listening room — terrain drawn from live sound
Drop a song · watch it become land
Each ridge below is one breath of the spectrum — bass at the left, air at the right, the last few seconds receding toward the horizon. Play something and watch the range rise.
Nothing playing — the terrain is idling on synthetic waves, not sound.
Your audio never leaves this tab. Files are decoded by your browser on this device — nothing is uploaded, to anyone, ever.
What the mountains mean, where your file goes (nowhere), and what the built-in track actually is.
An AnalyserNode reads the playing audio's frequency spectrum many times a second. Every 120 ms one snapshot — 88 log-spaced bands from deep bass to high air — becomes the newest ridge, and the older ridges step back toward the horizon. It is a slow spectrogram you can read like a landscape.
When you open a file, the page hands it straight to your browser's own audio decoder through a local object URL. There is no server in this loop and no network request carrying your music — you can watch the network panel of your browser's dev tools and see for yourself.
The built-in track is not a recording. It is generated live by Web Audio oscillators — a slow D-minor swell, a sub tone, filtered noise for surf, and sparse plucks through a delay — so it never repeats and contains no copyrighted audio. It exists so the terrain has something honest to measure if you have no file at hand.
When nothing plays, the terrain idles. The resting motion is a synthetic murmur — layered sine waves, clearly not sound — and the status line always says which of the two you are seeing. Nothing here pretends silence has a spectrum.