How to use it
Pick a source at the top. Use microphone asks for permission and draws whatever the microphone hears. Use computer audio analyses what your computer is playing, such as music, a video or a game, through the browser’s screen-sharing dialog (details in the questions below). Neither source is recorded or played back through the speakers, and the picture of your screen is not used. Open audio file decodes a recording locally and plays it. Play demo starts one of seven built-in test sounds.
Three views update together. The oscilloscope shows the waveform, the spectrum shows which frequencies are present right now, and the spectrogram shows how the spectrum changed over the last 16 seconds. On a phone, switch between them with the tabs.
Controls
- Freeze (Space or F) holds the picture; Save PNG downloads all three views as one image.
- Volume and Mute (M) change only what you hear from demos and files, never the analysis.
- Microphone input appears when the browser reports several audio inputs; your choice is remembered.
- FFT size, Smoothing (averages successive spectra), Input gain (scales the signal before analysis) and Display range (the decibel span of spectrum and spectrogram).
- Time base sets how many milliseconds fit across the oscilloscope. Trigger starts every trace at a rising zero crossing, so a steady tone stands still.
- Peak hold keeps the highest level seen at each frequency (switch it off and on to clear). Harmonic markers draw lines at f, 2f, 3f… of the detected pitch.
- Point at the spectrum or spectrogram, or drag a finger over it, to read a frequency. With the keyboard, focus the view and use the arrow keys; Escape removes the cursor.
How it works
Sound is a rapid variation of air pressure. A microphone turns it into a voltage, and the computer measures that voltage typically 44,100 or 48,000 times per second. The waveform on the oscilloscope is simply those samples plotted against time.
Any signal can be written as a sum of sine waves. The fast Fourier transform (FFT) takes a short block of N samples and works out how strong each frequency is. Its result is a row of N/2 frequency bins spaced by
At 48,000 Hz and an FFT size of 4096 the bins are 11.7 Hz apart and the block lasts 4096 ÷ 48,000 = 85 ms. A larger FFT separates close frequencies better, but its longer block blurs quick events.
A musical note is rarely one sine wave. It has a fundamental frequency f plus harmonics at 2f, 3f and so on, and their relative strengths give the instrument its timbre.
A spectrogram lays successive spectra side by side: time runs from left to right, frequency goes up and colour shows the level.
The pitch detector compares the waveform with a delayed copy of itself (normalised autocorrelation). The shortest delay at which the two match well is the period T, and the pitch is 1/T. If the match is poor, as with noise, the display shows a dash. The range is 40 Hz to 5 kHz, and notes are named relative to A4 = 440 Hz.
Levels are in decibels relative to full scale:
0 dBFS is the largest value digital audio can hold, −6 dB is half the amplitude and −20 dB is one tenth. The spectrum is calibrated so that a full-scale sine wave reads 0 dBFS. CLIP lights up when the signal reaches full scale.
Things to try
- Whistle a steady note. It is almost a pure sine wave: one tall peak. Slide the pitch and watch the line bend in the spectrogram.
- Sing “aah”, then change to “eee” on the same note. The harmonics stay where they are, but different ones become strong as your mouth changes shape.
- Clap. A click is short in time and spread over all frequencies, a vertical stripe in the spectrogram.
- Capture computer audio and play a song: bass and drums fill the left of the spectrum, cymbals the right.
- Compare the square and sawtooth demos: the square wave has only odd harmonics, the sawtooth has all of them.
Frequently asked questions
Is my audio recorded or sent anywhere?
No. The samples are read and drawn by this page in your browser and then discarded. During screen sharing only the sound is read; the picture is never used.
How do I capture the sound my computer is playing?
In Chrome or Edge on Windows, press Use computer audio, choose Entire screen in the sharing dialog and switch on the option to share system audio. Sharing a browser tab together with its audio also works, including on other desktop systems. A single application window usually gives no sound. Phone browsers do not offer this, and at the time of writing Firefox and Safari share no audio. Keep this page visible in its own window: it pauses while hidden.
Is there a way without screen sharing?
On Windows, some sound drivers offer a recording device called Stereo Mix. Enable it in the sound settings, allow the microphone here once and choose it in the Microphone input list. The visualizer then reads the computer’s output directly.
Why does the pitch show a dash?
The sound has no clear repeating pattern (noise, a clap, several unrelated notes), is too quiet, or lies outside 40 Hz to 5 kHz.