What it does
Every sound is a mix of pure tones. This analyser splits a sound into its frequencies with a fast Fourier transform (FFT) and shows:
- the spectrum: how loud each frequency is, in dBFS (decibels below the loudest level the recording can hold);
- a spectrogram: frequency up, time across, brighter for louder, so you can watch a whistle rise or a note decay;
- the strongest peaks, refined between FFT bins to a fraction of a hertz, with the nearest musical note and how many cents sharp or flat it is.
It works live from the microphone or on an audio file (WAV, MP3, OGG, FLAC and others your browser can decode). For a file, the spectra of overlapping slices are averaged (Welch’s method), which gives a steady, low-noise picture. The page never records, saves or uploads the sound: it is analysed in the browser and thrown away. There are no adverts on this tool.
How to use it
- Press Start microphone and allow it when the browser asks, or open an audio file, or press Try a test sound.
- Choose the FFT size. Bigger sizes separate close frequencies better (the resolution is the sample rate divided by the size) but follow changes more slowly.
- Choose a window: Hann suits most sounds; Blackman keeps a loud tone from hiding a quiet one next to it; rectangular gives the sharpest peak for a tone that fits the frame exactly.
- Set the frequency axis (logarithmic is how we hear), the lowest level shown, the live smoothing and peak hold.
- Read the peaks table. Freeze holds the live picture; Stop switches the microphone off. Download the spectrum as CSV, or the report as an image, PDF or Word file.
Worked example
Press Try a test sound: a made-up 2-second sound at 44.1 kHz with a 1 kHz tone at half of full scale and a quieter 3.15 kHz tone at a twentieth. With the default 4096-point Hann window the page averages 42 frames at a resolution of 10.8 Hz and shows Peak 999.9 Hz (B5 +21 cents), −6.0 dBFS. The second peak is 3.1499 kHz at −25.8 dBFS, a G7. The overall level is −9.0 dBFS. A tone at half of full scale has a peak of 20·lg 0.5 = −6.0 dBFS, as shown. Switch to 16384 points and the resolution becomes 2.69 Hz and the peak reads 1000.0 Hz. Change the tuning to A4 = 432 Hz and the same tone becomes C6 −47 cents.
Tips
- Phone and laptop microphones are not calibrated and often cut below about 100 Hz, so treat levels as relative. For sound pressure levels use a calibrated sound level meter.
- To tune an instrument, play one steady note and read the top peak; the cents show how far it is from the nearest note.
- Hum at 50 or 60 Hz and their multiples is mains interference.
FAQ
Is my sound recorded or sent anywhere? No. The microphone samples go straight into the analysis in this page and are discarded; nothing is saved or uploaded, and the page works offline.
Why does the browser say the microphone is switched off for this site? The site can block the microphone for all pages. You can still open an audio file.
Why is the peak not exactly the frequency I played? The FFT measures in steps (bins); the page fits a parabola across the top three bins, which is accurate to a small fraction of a bin for a steady tone. Use a bigger FFT size for more precision.
Where the numbers come from
- Fast Fourier transform (Cooley and Tukey, 1965), Hann, Hamming and Blackman windows, Welch's averaged periodogram (Welch, 1967) and parabolic peak interpolation: standard signal processing, written by fibub.
- Equal-tempered note names from f = 440 · 2^((n − 69)/12) (A4 adjustable).
- Checked against NumPy (numpy.fft.rfft) and SciPy (scipy.signal.get_window and scipy.signal.welch), and against tones of known frequency. The test sound is made up.
Standard scientific values, such as constants, atomic weights and melting points, are facts found in every textbook; we name where ours were checked. If you are an author or publisher and want a value, a credit or a table changed or removed, tell us and we will act quickly.
Last updated 2026-11-09.