Why is the curve so smooth compared with a live analyser?

Because frames across the whole file are averaged into one spectrum. A live analyser updates many times a second and keeps its ragged detail; this one makes tonal balance stable and readable at the cost of hiding momentary events.

What does the spectral centroid tell me?

It is the amplitude-weighted mean frequency, so it behaves like a brightness reading. Move it up and the material sounds more brilliant or noisier; move it down and the weight has shifted toward the low end. It is a summary, not a target.

Why is the peak frequency not the note I hear?

Because it is the loudest single bin in the average. In broadband music that bin can fall anywhere, and even in tonal material a strong harmonic may be louder than the fundamental. Trust it most on steady, clearly pitched signals.

What is the Hann window for?

The transform assumes each block is periodic, so cutting a block out of continuous audio creates edge discontinuities that smear energy across neighbouring bins. A Hann window tapers the block ends to suppress that smear, and its amplitude correction keeps the displayed level accurate.

Can I change the frequency resolution?

No. The analysis size is fixed at 8192 points, which sets both the reported bin width and the low-frequency detail. That choice favours reading the low end of a recording over fast time resolution, which is why the spectrogram exists for time-varying detail.

Spectrum Analyzer

You Might Also Need

Frequency Analyzer

Detailed frequency analysis

Spectrogram Viewer

View time-frequency spectrograms

Waveform Visualizer

Visualize audio waveforms in real-time

Goniometer / Vectorscope

Visual stereo image analysis

Spectrum Analyzer averages the whole file into one magnitude spectrum, so the frequency balance of a recording becomes a single readable curve. Peak frequency and spectral centroid are reported alongside it as numbers.

What is Spectrum Analyzer?

The canvas plots level in decibels against frequency, with zero decibels at the top and minus one hundred at the bottom, and grey gridlines every two kilohertz up to Nyquist so you can read a feature's place on the axis.

What the tool computes and shows:

The averaged frames count, the bin width in hertz, and the fixed display range of 0 dB at the top down to −100 dB at the bottom.
Peak frequency, the single bin with the most energy in the average, which is often the fundamental of a sustained tone.
Spectral centroid, the amplitude-weighted mean of the spectrum, which behaves like a brightness readout and moves up as high-frequency content increases.
Duration, taken from the decoded audio rather than from any metadata.

Analysis runs a windowed discrete Fourier transform at a size of 8192 samples per frame with a Hann window, and applies the amplitude correction that a Hann window requires so the plotted level stays honest instead of sinking by the window's gain factor. Frames are averaged across the file rather than shown one at a time, which trades time resolution for stability: broad tonal balance becomes clear and momentary events average out. Because 8192 points are analysed, low frequencies are resolved finely — useful for spotting a rumble shelf or a room resonance around fifty hertz — while the average smooths the ragged detail you would see in a live analyser. The spectrum is computed on the decoded channels in the page, so no upload and no encoder fetch is involved. The result is a magnitude spectrum averaged over the file and expressed in decibels against frequency, which is why the curve reads as a tonal balance rather than as a moment in time. Because the transform size is fixed, the reported bin width is fixed too, so every feature on the plot can be located to within one bin of the axis. Reading a magnitude spectrum well means remembering that one axis is logarithmic and the other linear, so equal distances across the plot do not correspond to equal musical intervals, and a bass note occupies far less of the width than its perceived pitch range would suggest.

Where the average misleads:

A short event such as a click or a single cymbal hit is diluted by the averaging and may not appear at all.
The curve is drawn per channel but only the first channel's curve is plotted, so a stereo file with very different sides shows one of them.
Peak frequency reports the loudest bin in the average, which for broadband music is a weak indicator compared with a tonal signal.
Bin width falls out of the analysis size, so the reported frequency resolution is fixed rather than adjustable.
The display clamps at −100 dB, so an extremely quiet floor is flattened against the bottom of the plot.

Use it to characterise material rather than to diagnose a moment: it tells you where the energy sits, whether the top end is rolled off, and whether a hum or rumble is a real part of the average. For a feature that changes over time, the spectrogram shows the same information with time on the horizontal axis.

How to use Spectrum Analyzer

Drop an audio file onto the panel; decoding and the transform run automatically.
Read the curve against the two kilohertz gridlines to see where the energy concentrates.
Take the peak frequency and spectral centroid cards as the tonal and brightness summaries.
Check the frames and bin width caption to understand how much averaging sits behind the curve.

When to use Spectrum Analyzer vs related tools

Spectrum Analyzer is the tool for the frequency balance of a whole file, answering whether the top end rolls off and where the weight sits. Spectrogram Viewer shows the same energy but across time, and Frequency Analyzer tracks a changing dominant pitch instead of averaging it away. When the question is level rather than frequency, RMS Meter and Loudness Meter (LUFS) give the numbers that matter.

Privacy & Security

This tool runs entirely in your browser — no data ever leaves your device. There is no server round-trip, no upload, no logging, and no account required. Your input is processed locally using client-side JavaScript and is never stored, transmitted, or accessible to anyone else. When you close the tab, everything disappears.