Logarithmic is almost always the right scale
Hearing is logarithmic: each octave doubles the frequency, so the octave from 100 to 200 Hz occupies the same musical distance as 5 kHz to 10 kHz. On a linear scale the entire bass register is squashed into the leftmost sliver while the top octave — which contains almost no musical information — takes half the screen.
Switch to linear only when you are hunting a specific high-frequency artefact, like switching-supply whine or aliasing.
Three views, three jobs
Bars show the balance right now — which frequency ranges are loud, whether a mix is bass-heavy, where a resonance sits. Waveform shows the signal in time, which is how you spot clipping and see the shape of a transient. Spectrogram scrolls, so you can watch frequency change over time — the view where a vibrato, a siren or a chirp becomes visible as a shape.
Smoothing is display, not data
Smoothing averages each frame with the previous ones. High values look calm and readable but hide short events; zero shows every frame exactly and flickers hard. Around 0.7 is a comfortable default for watching music, and near zero is what you want for catching a click or a pop.
The loudest-bin readout names the frequency as a musical note, which turns the analyser into a rough tuner and makes it easy to identify a hum: 50 or 60 Hz mains hum shows up immediately as a fixed spike.
Frequently asked questions
Why does it need microphone access?
The audio is analysed live in the page and never recorded or sent anywhere. Browsers also only allow microphone access over HTTPS, so the page must be loaded securely.
Is my audio uploaded?
No. The analysis happens in your browser — the file never leaves your device.
Something wrong with this tool, or an idea for it? Tell us