Spectrogram Viewer: See Any Sound Online

How to use the spectrogram viewer

1. Live: on the Microphone tab, press Start microphone and allow access. Sound scrolls in from the right. Pause freezes the picture; drag it or use the position slider to look back over the last two minutes, and Back to live returns to the scrolling view.
2. From a file: on the Audio file tab, drop a WAV, MP3, M4A, OGG, Opus or FLAC file, or a video for its soundtrack. The whole file appears at once. Zoom in to see detail, drag to move along it, and press Play to hear it with a playhead running across the picture.
3. Read it: point at any spot (or tap it on a phone) to read its frequency, the nearest musical note, the time and the level. The spectrum panel underneath shows the level of every frequency at that moment.
4. Keep it: Export PNG saves what is on screen with a frequency axis, a time axis and a colour scale. Share result copies a link that opens the viewer with your settings.

How to read a spectrogram

Time runs from left to right, frequency from bottom to top, and the colour of each point is how loud that frequency is at that moment. Bright means loud, dark means quiet.

• Thin horizontal lines are steady tones: a whistle, a test tone, the hum of a transformer. A tone from the tone generator played into the microphone draws a single line at its frequency, which is a quick way to check that the axis reads true.
• Stacks of evenly spaced lines are harmonics. Voices and most instruments produce a fundamental plus copies at two, three and four times its frequency. The lowest line is the pitch you hear; the spacing between the lines is the same number.
• Vertical stripes are short, sharp sounds such as claps, drum hits, clicks and consonants like "t" and "k". They cover many frequencies at once for a moment.
• Wide smears are noise: wind, fans, traffic, the "s" and "sh" of speech, tape hiss.
• A hard ceiling across a whole recording, with nothing above it, usually means a low-pass filter: lossy encoders such as MP3 remove the top of the spectrum to save bits. The About this file panel reports a sharp cut-off when it finds one.

Choosing the settings

Every spectrogram trades time detail against frequency detail. The FFT size sets the length of each analysis frame: the resolution line under the settings shows what it means at the current sample rate. At 48 kHz, a 512-point FFT uses 10.7 ms frames with bins 93.8 Hz apart; 8192 points use 171 ms frames with bins 5.9 Hz apart. Short frames show fast trills and drum hits crisply, long frames separate two tones a few hertz apart.

• Overlap makes frames share samples, so the picture gets more columns and smoother movement. 75% is a good default; 87.5% for slow scrolling and close zoom.
• Window: Hann suits almost everything. Blackman-Harris keeps a loud tone from leaking into its neighbours, which helps when you look for a quiet sound next to a loud one. Rectangular gives the sharpest peaks but the most leakage.
• Frequency axis: logarithmic for music and speech (every octave the same height), linear for bird song, the top end of a recording and encoder cut-offs, mel for speech and machine-learning work.
• Floor and Top set which levels map to the ends of the colour scale. If the picture is a dark field, lower the floor; if it's a bright blur, raise it. Auto levels picks both from what is on screen.

Tips and troubleshooting

• The microphone shows a flat dark field? Check the input with the microphone test first: the wrong device, a muted headset or a blocked permission all look like silence.
• Phone and laptop microphones, and Bluetooth headsets in call mode, often have a narrower range of their own. If the live view goes dark well below the top while a file of the same sound doesn't, the limit is the microphone, not the sound.
• A file shows nothing above 22 kHz? A 44.1 kHz file can't hold more than 22.05 kHz, and a 48 kHz file 24 kHz: the top of the picture is always half the sample rate.
• Very long files are analysed for their first 12 minutes or so (2²⁵ samples), and long files use a coarser time step to fit in memory; the resolution line says when that happens.
• Want the notes as MIDI rather than a picture? The audio to MIDI converter turns a melody into notes you can edit. For test tones across the audible range, try the hearing test.