Spectrum Analyzer

See live frequency bars and a spectrogram of your microphone input.

Everything runs in your browser. No audio is ever uploaded.

Analyze live audio

Allow microphone access so the analyzer can listen. The audio is analyzed on your device and is never recorded or uploaded.

Capture stops as soon as you leave or close this page.

Whistle, play a note or let a room hum: the tallest bar and the brightest line mark the frequency carrying the most energy. Bands are spaced logarithmically, so every octave gets about the same width.

This tool analyzes live microphone input only. Loading an audio file to analyze is not built yet.

A spectrum analyzer shows which frequencies a sound is made of instead of how loud it is overall. This free audio FFT analyzer, also called a real time frequency analyzer or an online spectrogram, listens through your microphone and draws either frequency bars or a scrolling picture of the last few seconds.

How to use it

Press Allow microphone and accept the browser prompt. The audio is analyzed on your device and never recorded or uploaded, and capture stops when you leave the page.
Choose a display. Frequency bars show the spectrum at this instant, which is what you want for a steady sound. The spectrogram scrolls time from left to right and is better for anything that changes.
Make a sound. Whistle, play a note, hum, or simply let the room hum on its own. The tallest bar and the brightest line mark the frequency carrying the most energy.
Read the two figures. Loudest frequency gives the peak in hertz with its nearest musical note, and input level shows how hot the signal is in dBFS so you know whether the reading is trustworthy.
Press Freeze to hold the picture while you look at it. The microphone stays on, so Resume picks straight back up. The note at the bottom tells you how many hertz one analysis bin covers.

What you can use it for

Identifying a mystery hum. Hold the analyzer near the noise and read the peak: a strong line at 50 or 60 Hz and its multiples is mains related, while a fan or a pump usually sits higher and wanders.

Checking what a speaker actually produces. Play a tone or a sweep on one device and watch it here on another. You will see where the output rolls off, and whether harmonics appear that were never in the signal.

Practising an instrument or a voice. The peak readout names the nearest note, and the spectrogram shows vibrato, drift and the harmonics above the fundamental as they happen.

Learning what frequency bands sound like. Play music, watch which bands move with the kick drum, the voice and the cymbals, and the words bass, midrange and presence stop being abstract.

Finding a room resonance. Play a slow sweep from another device and watch for a band that jumps out of line. Freeze it, note the frequency, and move the microphone to see whether it follows the room or the speaker.

The frequency bands and what lives in them

Band	Range	Typical sources
Sub-bass	20 to 60 Hz	Pipe organ pedals, cinema rumble, the lowest notes of a five string bass. Felt more than heard, and invisible to most microphones in a laptop.
Bass	60 to 250 Hz	Kick drum, bass guitar, the fundamentals of a male speaking voice, mains hum and its first harmonics.
Low midrange	250 to 500 Hz	The body of guitars, pianos and most voices. Too much here is what people call boxy or muddy.
Midrange	500 Hz to 2 kHz	The core of speech and the working range of nearly every instrument. A telephone keeps this band and throws away the rest.
Upper midrange	2 to 4 kHz	Consonants, attack and bite. Hearing is at its most sensitive here, so a small change is easy to notice.
Presence	4 to 6 kHz	Clarity and definition. This band decides whether a voice sounds close or distant.
Brilliance	6 to 20 kHz	Cymbals, sibilance, birdsong, air. The top of it disappears with age for every listener, and cheap microphones give up there too.

Accuracy and limits

The analysis window is fixed at 2048 samples, so one bin covers roughly 20 Hz at a common sample rate. That is plenty in the treble, where an octave spans thousands of hertz, and coarse in the bass, where a whole octave can fall into a handful of bins. Treat a low frequency reading as approximate.

You are looking at your microphone as much as at the sound. Laptop and phone microphones roll off the bass sharply, have peaks of their own in the upper midrange, and the browser may add noise suppression or automatic gain on top. A shape on screen is not a measurement of the source.

The vertical scale is relative. It shows which frequencies carry more energy than the others right now, not absolute levels in decibels. For a level figure, however approximate, use the decibel meter instead.

Frequently asked questions

What does a spectrum analyzer show?

How the energy of a sound is spread across frequency. A single whistled note gives one tall bar plus smaller harmonics above it; speech gives a moving cluster in the midrange; broadband noise fills the whole width at a similar height.

What is the difference between bars and a spectrogram?

Bars show one moment: frequency across the bottom, energy as height. A spectrogram shows a history: time from left to right, frequency from bottom to top, energy as colour. Use bars for steady sounds and the spectrogram for anything that moves.

Why does the peak frequency jump around?

Because real sounds are not single tones. Two bins can be nearly equal and swap places between frames, especially with background noise present. Get closer to the source, make the sound steadier, or freeze the display and read the shape rather than the number.

Can I analyze an audio file instead of the microphone?

Not yet. This page analyzes live microphone input only. The workaround is to play the file on the same or another device and let the analyzer listen, which is fine for finding frequencies though the room and the microphone are then part of what you see.

Why are the low bands so wide on screen?

The display is spaced logarithmically, so every octave gets about the same width. That matches how hearing works: the octave from 100 to 200 Hz is as musically large as the one from 5 to 10 kHz, even though it covers a hundredth of the hertz.

Is a browser spectrum analyzer accurate enough to be useful?

For identifying frequencies, comparing before and after, and understanding what a sound is made of, yes. For calibrated measurements of a speaker or a room, no, because that needs a measurement microphone with a known response and software that knows its calibration.