Spectrogram (waterfall), live in your browser
Watch sound as a scrolling colour picture, live from your mic, a file or a browser tab.
Press Listen to see your sound as a picture
The browser asks to use your microphone. Your sound stays on this device.
Or drop an audio file here.
Picture
Raise “Dark at” to hide room noise; lower “Bright at” to make soft sounds glow.
Detail
Frequency
Record while showing
Nothing recorded yet.
The sound is kept in this browser tab only, up to 10 minutes, and saved as a file on your device when you press Save WAV.
Source
Files are read on this device only; they are never uploaded.
Export and share
CSV lists the strongest frequency over time. The link keeps your settings, never your sound.
Recent snapshots
How to use the spectrogram
- Press Listen and allow the microphone. No mic? Pick Audio file or Demo sound as the source, or Tab audio on a computer.
- Pick a preset: Voice, Music, Birds or Full range. Each one sets the detail, the speed and the frequency scale for that job.
- Watch the picture scroll. Point at it with the mouse, or press and hold on a phone, to read the time, frequency and level at that spot.
- Press Pause to stop the picture. Drag it, or move the Look back slider, to go back up to 60 seconds.
- Press Snapshot to save what you see as a PNG, or Full session for everything up to 10 minutes. Turn on Record while showing to keep the sound as a WAV file too.
How to read a spectrogram
Time runs sideways
The newest sound appears at the right edge and slides left as it gets older. The labels under the picture count back in seconds, so “−4 s” happened four seconds ago. In the top-to-bottom layout, often called a waterfall, the newest line is at the top and time flows down.
Height is frequency
Low sounds sit near the bottom and high sounds near the top. On the log scale every octave gets the same height, so 100 to 200 Hz takes as much room as 5 to 10 kHz, which is close to how we hear. The linear scale gives the same room to every hertz and suits whistles and high harmonics. The mel scale sits between the two and is common in speech work.
Colour is loudness
Dark is quiet and bright is loud. With the Ember colours the order runs dark blue, ice blue, amber, coral and almost white. The bar under the picture shows the level at each end, for example −100 dB for the darkest colour and 0 dB for the brightest. Raise the dark level to hide room noise, or lower the bright level to make soft sounds glow.
Use cases
See bird song
Most songbirds sing between about 2 and 8 kHz, in notes that last a tenth of a second or less. The Birds preset shows 1 to 12 kHz on a linear scale with FFT 1,024, so each chirp appears as a short hook or slash and a trill as a row of tiny dashes. Pause and look back to compare two calls side by side.
Spot a buzz that comes and goes
A fridge, a light dimmer or a ground loop can add a buzz that starts and stops. Here it shows as a flat line, often at 50 or 60 Hz and its multiples, that switches on and off with the device. Leave the picture running at 10 seconds per screen, then look back to see when it began. To read the exact hum frequency, use the live spectrum analyzer.
Look at speech
Say a slow “ee, ah, oo”. Voiced sounds show as stacks of thin lines, the harmonics, that rise and fall with your pitch, plus brighter bands where your mouth shapes the vowel. Hissing sounds such as “s” and “sh” show as haze from about 4 to 10 kHz. The Voice preset uses FFT 2,048 on a mel scale, a good balance of timing and pitch detail for talking.
Check a recording for noise
Open the file, choose the linear scale and watch the top of the picture. Steady hiss is an even haze across the whole width, a thin line near 15.7 kHz can be whine from old electronics, and a hard edge at 16 kHz means the file went through a lossy encoder such as a 128 kbps MP3. For one picture of a whole file with zoom, make a spectrogram of a whole file.
Tips for a sharp picture
The FFT size sets the trade between timing and pitch detail. A small size such as 512 or 1,024 covers only 11 to 21 ms of sound at 48 kHz, so clicks and drum hits stay sharp, but notes close together blur into one band. A large size such as 8,192 or 16,384 splits notes about 10 Hz apart, but smears fast events over a sixth of a second or more. Start with 2,048 for speech and 8,192 for music.
Overlap sets how much each slice shares with the next. At 50 % the picture gets fewer columns and looks blocky; at 87.5 % a new column comes eight times per FFT length, so movement looks smooth. If a slow phone cannot keep up, the tool lowers the column rate by itself and says so.
Set the contrast last. Put “Dark at” just under the room noise, which is often around −90 to −100 dB here, and the sounds you care about will stand out. To see why this trade exists, read how the FFT works.
Spectrogram vs spectrum analyzer
A spectrum analyzer shows one moment: frequency from left to right and level up and down, many times a second. It is the best way to read an exact peak, such as 1,000.0 Hz at −6 dB. A spectrogram keeps the history. It lines up thousands of those moments side by side and turns level into colour, so you see how the sound changes: a note that slides, a hum that switches on, a call that repeats every 3 seconds.
Use the spectrogram to find and follow a sound, then measure it on the live spectrum analyzer. For the tone balance of a speaker or a room, the 1/3-octave view is easier to read than either. More audio analyzer tools are in the same group.
FAQ
What is a spectrogram?
A picture of sound over time: left to right is time, bottom to top is frequency, and brighter colour means louder.
How do I read a spectrogram?
Horizontal lines are steady tones, stacked lines are harmonics of one note, vertical stripes are clicks or drum hits, and haze is noise.
Why does it look blurry?
Small FFT sizes give sharp timing but blurry frequency; large sizes do the opposite. Try 2,048 for speech and 8,192 for music.
Can I save the spectrogram?
Yes. Save the visible picture as PNG, or the whole session up to 10 minutes. Turn on “Record while showing” to keep the sound as a WAV file as well.
Can I make a spectrogram of a file?
Yes. Choose “Audio file”; for a full-file picture with zoom use the Audio File Spectrum Analyzer.