Loading tool…
A waveform tells you how loud a recording is; a spectrogram shows what it is made of. Open a file or switch on your microphone to see frequency against time, hover any point for an exact reading, and save the picture as a PNG.
Processing locally on your device
Overview
A spectrogram is built from a long series of fast Fourier transforms (FFTs). The audio is cut into short, overlapping windows, each window is shaped with a Hann window to reduce spectral leakage, and the FFT turns it into a spectrum: how much energy sits at each frequency. Drawing those spectra side by side, with colour standing for level in decibels, gives a picture where time runs left to right and pitch runs bottom to top. For files, this tool runs its own radix-2 FFT over the decoded samples in your browser, up to 1,200 columns across the whole recording, and calibrates the scale so a full-scale sine wave reads about 0 dB.
FFT size controls the classic trade-off between frequency detail and timing detail. Frequency resolution is the sample rate divided by the FFT size, so at 48 kHz an FFT of 512 gives bins about 94 Hz wide and a window of roughly 11 ms, while 8192 resolves about 5.9 Hz but smears timing over about 170 ms. Small sizes suit drums, consonants and clicks; large sizes separate musical notes, harmonics and narrow hum lines. The logarithmic axis gives each octave the same height, which matches how we hear and keeps bass detail visible, while the linear axis is better for spotting evenly spaced harmonics and high-frequency cut-offs.
The floor and ceiling sliders set which levels map to the darkest and brightest colours. Lower the floor towards −140 dB to reveal faint noise and reverb tails, or raise it to −70 dB to show only the dominant content. Five colour maps are included: Magma, Viridis and Inferno are perceptually uniform, Classic mimics older analysers, and Grayscale prints well. A top-frequency limit (16, 8, 4 or 2 kHz) zooms into the range you care about, and stereo files can be viewed as a mix or as the left or right channel alone.
Spectrograms are a diagnostic tool. A bright horizontal line at 50 or 60 Hz with copies at multiples reveals mains hum. A shelf where everything stops at 16 kHz suggests the file once passed through a low-bitrate MP3, even if it is now a WAV or FLAC. Speech shows formant bands, birdsong appears as sweeps and chirps, and a microphone check in live mode shows instantly whether the mic picks up fan noise or room resonances.
| FFT size | Bin width | Window length | Best for |
|---|---|---|---|
| 512 | ≈ 94 Hz | ≈ 11 ms | Drums, clicks, speech timing |
| 1024 | ≈ 47 Hz | ≈ 21 ms | Speech, general overview |
| 2048 | ≈ 23 Hz | ≈ 43 ms | Music overview (default) |
| 4096 | ≈ 12 Hz | ≈ 85 ms | Musical notes, harmonics |
| 8192 | ≈ 5.9 Hz | ≈ 171 ms | Hum lines, fine pitch detail |
Drop an audio or video file onto the page, or switch the source to Microphone and press Start.
Choose an FFT size: 1024–2048 for general use, 4096–8192 for musical pitch, 512 for transients.
Pick a logarithmic or linear frequency axis, a top frequency and a colour map.
Adjust the dB floor and ceiling until the detail you need stands out.
Hover the image to read time, frequency, note and level; click to play from that point.
Press Export PNG to save the spectrogram with its axes and colour scale.
Use cases
Spot mains hum, whine from a laptop charger, air-conditioning rumble or a single click before you reach for a noise-reduction tool, so you know exactly which frequencies to target.
A genuine CD-quality file usually shows content up to 20 kHz or more. A hard ceiling at 16 kHz or 19 kHz is a common sign that it was transcoded from a lossy MP3 or AAC.
Singers and instrumentalists can see vibrato, harmonics and the attack of each note, and speech students can compare vowel formants between recordings.
Bird calls, bat detector recordings and other wildlife sounds are easier to identify and count as shapes on a spectrogram than by ear alone.
Pro tips
Answers to common questions about the spectrogram.
Colour shows level in decibels relative to full scale. With the default range, the brightest colour is about 0 dB (the loudest possible signal) and black is −110 dB or quieter. The colour bar on the right of the image is labelled so you can read values directly, and the hover readout gives the exact figure.
Use 2048 as a starting point. Choose 4096 or 8192 when you need to separate close frequencies such as musical notes or hum harmonics, and 512 or 1024 when you need precise timing, such as drum hits or clicks. Larger sizes are a little slower to compute.
Live mode uses the browser’s AnalyserNode, which applies a Blackman window and draws one column per screen refresh, while files use a Hann window and a fixed number of columns. Levels in live mode are corrected to approximate dBFS, but small differences of a few decibels are normal.
Yes. Drop a video file and the soundtrack is decoded and analysed. The video picture is ignored.
There is no fixed limit, but the whole file is decoded into memory, which takes about 23 MB per minute of stereo audio at 48 kHz. Files over 30 minutes work on most computers, though each column then represents several seconds of audio.
No. ToolEnhancer processes files on your own device inside the browser. They are never sent to, stored on or seen by our servers.
Yes. The spectrogram is free to use with no sign-up, no watermarks and no daily limits.