Audio Spectrum Analyzer & Spectrogram

Any sound — a voice, a violin, a fan — is a single wiggling pressure wave. Yet your ear effortlessly hears it as a blend of pitches: a low rumble here, a bright hiss there. The mathematical tool that pulls a complicated wave apart into the individual frequencies it is made of is the Fourier transform, and the fast version your browser runs thousands of times a second is the FFT (Fast Fourier Transform). This analyzer feeds live audio into the Web Audio API’s FFT and paints the result.

The spectrum (the top graph) is one FFT frame: the horizontal axis is frequency, low on the left and high on the right, and the height of each point is how much energy the sound has at that frequency right now. A pure tone is a single spike; a rich instrument is a whole comb of spikes (its harmonics); noise is a broad smear. The spectrogram (the bottom graph) stacks those frames over time — each new vertical stripe is the latest spectrum, brightness shows energy, and as fresh stripes appear on the right the older ones scroll left, so you literally watch the sound’s frequencies evolve.

There is no free lunch. A larger FFT size chops the spectrum into more, narrower bins, so you can tell two nearby notes apart — but it needs more samples to do it, which means it reacts more slowly and blurs fast changes. A smaller FFT size is nimble in time but coarse in frequency. This is the time–frequency trade-off, and choosing where to sit on it is the whole art of spectral analysis. The smoothing control averages each bin with its recent past to calm the flicker, at the cost of a little responsiveness.

By default the analyzer draws frequency on a logarithmic axis, because that is how we hear. Each musical octave doubles the frequency — 110, 220, 440, 880 Hz are all the note A — so on a log axis those octaves are evenly spaced, matching a piano keyboard. On a linear axis the entire bottom four octaves would be crammed into the leftmost sliver while the top, where we hear little musical detail, would hog the rest. Switch between the two and watch the low notes spread out or bunch up. The peak-note readout uses the equal-tempered formula above and reports how many cents (hundredths of a semitone) the loudest frequency sits above or below the nearest note — a built-in tuner.

This is a complete spectrum analyzer, spectrogram, level meter, and chromatic tuner in one page, running entirely on your device through the Web Audio API. Microphone audio and any file you open are processed locally and never leave your browser — there is no upload, no recording, and no account. It works offline once the page has loaded.

Use it to hunt down a mystery hum, check whether your headphones really reproduce the low bass, tune a guitar by ear-and-eye, watch the harmonics that give each instrument its character, or simply learn what sound looks like. To go deeper on the maths and the musical pitch behind it, read Sound, Frequency & the FFT.