How can a spectrogram tell whether a FLAC is really lossless?

Load the file with the highest frequency set to “All”. Genuine CD-quality audio usually has content reaching about 20 kHz. A clean, flat cutoff around 16 kHz or 19 kHz with near-black above it is the typical fingerprint of a file transcoded from MP3 or AAC. Some recordings are naturally dull, so compare several tracks before drawing conclusions.

Which FFT size should I pick?

Bigger windows resolve pitch more finely but blur timing; smaller ones show sharp transients but smear frequency. Use 512–1024 for speech and drums, 2048–4096 for musical pitch and harmonics, and 8192 for low-frequency detail. More overlap gives a smoother image at the cost of extra computation.

Why does the axis go up to 24,000 Hz?

Browsers usually decode audio at 48 kHz, so the highest representable frequency (the Nyquist frequency) is 24 kHz. A 44.1 kHz file will be empty above 22.05 kHz, which is expected. Use the Highest frequency option (20k, 16k…) to show only the useful range.

Is my audio uploaded? Is it free?

Nothing is uploaded — decoding and the spectrum computation run entirely in your browser and are discarded when you close the page. The tool is free, with no sign-up, no limits and no watermark.

🌌

🌌Audio Spectrogram Online - STFT Frequency Analysis with Log/Linear Axis, Export PNG

A spectrogram draws sound as an image: time runs left to right, frequency bottom to top, and brighter colors mean more energy at that moment and pitch. It is the go-to way to spot “fake lossless” music (a hard cutoff at 16 kHz means it was converted from MP3), study vocal and instrument harmonics, find hum, feedback or noise frequencies in a recording, analyze birdsong and speech, or make striking artwork from audio. This tool decodes the file with Web Audio and computes a Hann-windowed short-time Fourier transform with a radix-2 FFT written in JavaScript, processed in chunks with a progress bar so long files stay responsive. Choose an FFT size from 512 to 8192 and 50–87.5% overlap, a logarithmic or linear frequency axis with custom limits, the dB floor and ceiling that set the contrast, and Magma, Viridis, Inferno or grayscale colors on a dark or light background. Click anywhere to read the time, frequency (with the nearest musical note, such as A4) and level in dB, or analyze just a segment to zoom in. Export a PNG with time axis, frequency axis and color bar.

Free · No sign-up · Runs in your browserLast updated: 2026-10-26

Loading tool...

How to use Audio Spectrogram

1Upload an audio file (MP3, WAV, FLAC, M4A, OGG…) or a video — it is decoded and plotted automatically
2Adjust FFT size, frequency axis, dB range and colors; click the image to read time / frequency / dB
3Optionally set a start and end time to analyze one section, then click Export PNG

FAQ

How can a spectrogram tell whether a FLAC is really lossless?: Load the file with the highest frequency set to “All”. Genuine CD-quality audio usually has content reaching about 20 kHz. A clean, flat cutoff around 16 kHz or 19 kHz with near-black above it is the typical fingerprint of a file transcoded from MP3 or AAC. Some recordings are naturally dull, so compare several tracks before drawing conclusions.
Which FFT size should I pick?: Bigger windows resolve pitch more finely but blur timing; smaller ones show sharp transients but smear frequency. Use 512–1024 for speech and drums, 2048–4096 for musical pitch and harmonics, and 8192 for low-frequency detail. More overlap gives a smoother image at the cost of extra computation.
Why does the axis go up to 24,000 Hz?: Browsers usually decode audio at 48 kHz, so the highest representable frequency (the Nyquist frequency) is 24 kHz. A 44.1 kHz file will be empty above 22.05 kHz, which is expected. Use the Highest frequency option (20k, 16k…) to show only the useful range.
Is my audio uploaded? Is it free?: Nothing is uploaded — decoding and the spectrum computation run entirely in your browser and are discarded when you close the page. The tool is free, with no sign-up, no limits and no watermark.