∿ Audio spectrogram viewer

How to use it

Drop an audio file onto the frame, or click it and pick one. Anything your browser can play will do: WAV, MP3, OGG, FLAC. A second later the spectrogram appears: time runs left to right, frequency bottom to top, and brightness is loudness. The first two minutes of the recording are shown.

The Top and Bottom sliders choose the band of frequencies. Anything drawn into a recording usually sits in a narrow band, and stretching that band over the whole plot makes it much easier to read. The “1.8–7.2 kHz” button jumps straight to the band this tool paints in. Contrast cuts away quiet noise: fewer decibels, cleaner background. Window, the slice size for the Fourier transform, trades one sharpness for another: a large window is precise in frequency, a small one in time. Play sounds the file while a line runs across the picture.

Underneath is the reverse. Type a word and the tool paints it in sound. You can listen to it, look at it, download it as WAV and send it up to the viewer.

How it works

The sound is cut into short, overlapping slices. Each slice is multiplied by a smooth Hann window so its edges don’t invent frequencies, then the fast Fourier transform splits it into sine waves. The loudness of each sine, in decibels, becomes one column of pixels. Lay the columns side by side and you have a spectrogram.

Painting runs the other way. Every row of a letter gets its own band of frequencies, every column its own slice of time. Where a pixel is lit, sines from that band sound; where it is dark, there is silence. Soft edges on each slice keep the letters from smearing. The file is 16-bit mono at 22,050 Hz.

Questions

I loaded a song and see nothing. Is that right?

Usually, yes: music holds harmonics, drum hits and noise, not pictures. But push the top limit all the way up and look along the whole track: some musicians have hidden pictures in sound exactly this way. The steganography lesson has the best-known case.

Why are the letters in my file blurry?

MP3 compression throws away what the ear barely hears, quiet high frequencies included. Keep such files as WAV. A larger window, 2048 or 4096, also helps.

Can a picture be hidden so nobody hears it?

Not completely: the sines are audible. But they can be made quiet, moved above 15 kHz where many adults no longer hear, or laid under loud music.

Is my file uploaded anywhere?

No. The sound is analysed in your browser and goes nowhere.