Audio File Spectrogram
Draw a Hann-windowed spectrogram of an audio file, with a start and end time and a choice of FFT size.
🔒 Processed entirely in your browser — nothing you enter here is ever uploaded.
Result
About this tool
Open an audio file and the page decodes it, mixes every channel down to one mono buffer, and paints a spectrogram of the range you set. Time runs left to right. Frequency runs bottom to top. Color is the strength of an FFT bin, mapped so that levels from about −96 dB toward −12 dB climb through the heat colors. The caption reports the length of the window, the FFT size, and the loudest bin it found, with that bin’s frequency, level, and time.
FFT sizes are 512, 1024, 2048, 4096, and 8192. The hop is one quarter of the size, and never under 128 samples, so 2048 steps 512 samples between columns. Each column is one Hann-windowed frame. The end time is forced to at least 0.05 seconds after the start and cannot pass the file duration. The frequency axis can be linear or logarithmic, and a grid can be toggled. A larger FFT narrows each frequency bin and widens the time step, so a short click and a low hum will not both look sharp at the same setting.
Stereo is summed before the transform. A tone only on the left and the same tone only on the right paint the same column. The page does not listen to the microphone; that live view is the realtime spectrum analyzer. It also does not export a table of bins or a new audio file. The picture is the result, drawn from samples that never leave the browser.
Example. A file drawn with FFT size 2048 uses a hop of 512 samples and paints one canvas column per hop for the chosen start and end.
Limit. The result is a picture plus a peak label. The page does not export a spectrum table or a new audio file, and the end time must sit at least 0.05 seconds after the start.
How do I read an online spectrogram of my file?
Time runs left to right across the start and end you typed. Frequency runs bottom to top, on a linear or logarithmic scale. Brighter color is a stronger FFT bin. The caption names the loudest bin in that window, with its frequency, level, and time.
What FFT size and hop does 2048 use?
The size control accepts 512, 1024, 2048, 4096, and 8192. The hop is one quarter of that size, and never under 128 samples, so 2048 advances 512 samples per column. Each frame is multiplied by a Hann window before the transform.
Can I save the spectrogram as audio or a data table?
The page draws a canvas and a peak caption from the mono mix of the decoded file. A left-only tone and a right-only tone at the same frequency paint the same column. Save the picture from the browser if you need a still; the tool itself writes neither CSV nor a new audio file.