These are browser-playable renders from the current ClatterDrive engine. They are not
recordings of a real disk and they are not bundled sound effects. Each clip comes from
a different simulated workload and mounting condition so the low end, tick texture, and
enclosure bloom actually change.
Cold power-on through spindle run-up, early servo activity, a short idle period, then
a final park. This one uses the desk-coupled mounting profile.
Why it sounds this way: the desk/table structure path adds more low
body during run-up, while the late sharp transient is modeled park/contact behavior.
Active drive, park into a quieter standby-like state, then wake back into seek and
transfer activity. This is also rendered with the desk-coupled profile.
Why it sounds this way: the low-frequency mounting path stays present
through the wake cycle, so the main contrast is the park/wake transition rather than a
simple change in seek rate.
Dense bursts of metadata and journal-style seeks with occasional recalibration-like
interruptions and heavier flush-like jumps, rendered with the bright bare-drive lab
profile.
Why it sounds this way: this one strips away most desk/enclosure bloom
so the servo-wedge tick texture and brighter exposed seek structure are easier to hear.
The same repo that serves the fake drive also renders these clips. Regenerate them with
uv run python -m tools.generate_readme_demo_samples.
The current README and this page intentionally use three different situations rather than
three almost-identical steady-state loops, so the differences in structure coupling,
power-state behavior, and seek texture are actually audible.
These public demo clips are peak-normalized for browser playback. The normalization is
only for the checked-in sample artifacts; it does not change the live engine model.