Does the waveform on screen actually match the audio? This page draws a
3 s 440 Hz tone with a silent middle third, then reads back
individual canvas pixels to prove sound paints as waveform, silence stays
blank, and the loop overlay lands exactly on the 1.0–2.0 s window.
Technically: the sine is rendered via getWaveformPeaks →
renderWaveform (peaks/mirror), then addLoopRegions
overlays 1.0–2.0 s; badges probe actual canvas pixels for the
amplitude→pixel and time→x mappings. Below: analyzeWaveform
stats vs closed-form sine truth from a real AudioBuffer, the repaired
renderStereoWaveform, and the repaired idempotent HiDPI
resize.