What does a spectrogram show?

It shows which frequencies are in a sound and how loud they are, moment by moment: time runs from left to right, frequency from bottom to top, and colour shows the level. A steady tone draws a flat line, speech a stack of moving lines, and noise a haze.

What do the colours on the spectrogram mean?

Brighter means louder. The brightest colour is the loudest point in the file, and the colours fade to dark as the level drops, down to the dynamic range you set. Point at any spot to read its level in dB.

Which FFT size should I use?

It is a trade-off between frequency and time. A bigger FFT (up to 4096) separates close notes but smears quick sounds; a smaller one (down to 512) shows clicks and beats sharply but blurs pitch. 2048 is a good start; the table above lists the steps at 48 kHz.

Is the microphone sound recorded or uploaded?

No. Your browser asks for permission first, and the sound is analysed in your browser to draw the picture; it is not recorded, saved or uploaded. Stop turns the microphone off. The page records ordinary analytics events, such as the settings you chose, not the sound.

Which files does it take, and how long can they be?

MP3, M4A, AAC, WAV, OGG, OPUS, FLAC, WEBM and MP4 files up to 100 MB and 30 minutes. The file is decoded in your browser's memory and is not uploaded, which is why there is a limit.

Online Spectrogram

See the frequencies in an audio file or your microphone as a spectrogram: time runs left to right, pitch bottom to top, and louder sound glows brighter. Point at any spot to read its time, frequency and level, change the axis and colours, and save the picture as a PNG.

Drop an audio file here or click to browse

MP3, M4A, AAC, WAV, OGG, OPUS, FLAC, WEBM or MP4, up to 100 MB and 30 minutes.

Drop a file to draw its spectrogram here.

Point at or tap the image to read the time, frequency and level.

Settings

FFT size

Bigger shows finer frequency detail but blurs quick sounds; smaller is sharper in time but coarser in frequency.

Overlap

Frequency axis

Top frequency

Colours

Dynamic range

90 dB

Sounds this far below the loudest point fade into the background colour.

The file is processed in your browser and is not uploaded.

How to make a spectrogram online

Everything runs in your browser.

1
Drop a file or start the microphone
An MP3, M4A, WAV, FLAC or other common file up to 100 MB and 30 minutes, or the Microphone tab for a live, scrolling view of the last 20 seconds.
2
Read and adjust it
Point at or tap the image to read the time, the frequency and the level there. Switch to a log axis for speech and music, set the top frequency, the dynamic range, the FFT size and the colours.
3
Save it as a PNG
Download PNG saves the picture with its frequency and time axes, named after your file.

Want the loudness over time as a picture instead, for a post or a cover? The Audio Waveform Generator draws it as a PNG or SVG.

What the spectrogram shows

Every frequency over time
Each column is a short slice of the sound split into frequencies, so a steady tone is a flat line, a siren a curve and a drum hit a tall stripe.
Loudness as colour
The loudest point in the file gets the brightest colour; quieter sound fades to dark. The dynamic range slider sets how far down the colours reach.
Voices, harmonics and noise
A voice or an instrument shows a stack of parallel lines (the harmonics); hiss is a haze across the top, hum a line at the bottom.
Live from the microphone
The Microphone tab scrolls as you speak, play or whistle, which makes it easy to see how a sound is built.

When a spectrogram helps

Find a hum or hiss before cleaning a recording
Mains hum shows as a thin line low down, hiss as a haze at the top, a whine as a line high up. Knowing which one you have tells you what to fix.
Check a lossy file's top end
MP3 and other lossy encoders filter out the highest frequencies, more so at low bitrates. On the spectrogram that shows as a flat edge where the colour stops.
Birdsong, instruments and voices
See the shape of a bird's call, the harmonics of a note or the vowels in a word, and compare two recordings side by side.
Puzzles and classes
Some tracks hide a picture or text that only a spectrogram shows; a log axis and the Heat colours make it easier to read. It is also a quick demo for a physics or audio class.

Found hiss or hum? Clean it with the Audio Enhancer. Recording interviews: how to record an interview.

FFT size: frequency detail against time detail

At 48 kHz, each FFT size gives this frequency step and frame length:

FFT size	Frequency step	Frame length
512	93.8 Hz	10.7 ms
1,024	46.9 Hz	21.3 ms
2,048	23.4 Hz	42.7 ms
4,096	11.7 Hz	85.3 ms

Frequency step = sample rate ÷ FFT size; frame length = FFT size ÷ sample rate. Pick a big size to separate close notes, a small one to see quick clicks and beats.

What the colours mean

Colour shows level in decibels relative to the loudest point in the file: the brightest colour is that loudest point, and each step down the colour map is quieter. Below the dynamic range you set, sound fades into the background colour.

A small dynamic range (40 dB) shows only the strong parts, such as the main notes of a song. A large one (up to 120 dB) brings out quiet detail, such as room noise, hiss and the faint top of a recording.

How the picture is made

For a file, the page cuts the sound into overlapping frames, multiplies each by a Hann window, and turns it into frequencies with a fast Fourier transform. Long files are drawn with up to 4,000 columns, so the picture stays sharp without freezing the page.

The microphone view uses the browser's own analyser, which applies a Blackman window, so its colours are close to but not exactly the same as the file view's. The browser decodes a file at its own sample rate, so the top of the axis is half that rate.

When the recording is a meeting or an interview

A spectrogram shows how a recording sounds. When what you need is what was said, ParrotNotes records the meeting or interview on your phone and gives you the transcript and a summary.

Get Started Free

Related tools