
To check whether a FLAC or WAV is really lossless, drop it into a free online Spectrogram and look at where the high frequencies stop. A file made from an MP3 or AAC usually hits a hard, flat ceiling somewhere between about 16 and 20 kHz, while a genuine CD-quality file keeps going toward 22 kHz and thins out unevenly with the music. The check takes about a minute, and the result is strong evidence rather than proof.
Lossy encoders save space by discarding what you're least likely to hear, starting with the very top of the spectrum. Most MP3 and AAC encoders apply a low-pass filter first (a filter that lets lower frequencies through and blocks higher ones), so everything above a cutoff frequency is removed. The lower the bitrate, the lower that cutoff tends to be.
Converting that MP3 to FLAC or WAV doesn't put anything back. The new file stores the decoded MP3 perfectly, missing top end included, so the ceiling survives the conversion. That's why it's the first thing to look for when you turn audio into a spectrogram. If the difference between the two kinds of format is new to you, lossy vs lossless audio covers it, and what a FLAC file is explains why FLAC on its own guarantees nothing about where the audio came from.
Open the Spectrogram and drop the file in. It decodes in your browser, mixes the audio to mono and analyses it in a background worker, with "Analysing…" and a percentage while it runs. It's free, there's no account, and your audio never leaves your device.
Switch Frequency scale to Linear. The default Logarithmic view gives every octave the same height, so the whole 10–20 kHz octave gets squeezed into a thin band at the top. On Linear, the upper half of the image covers roughly 11 to 22 kHz, and a cutoff becomes a wide, obvious line instead of a sliver.
Set Window size to 4096 · finer pitch or 8192 · finest pitch. A larger window gives finer frequency detail, so the horizontal edge of a lossy cutoff comes out crisp instead of fuzzy. Timing gets blurrier, but you don't care about timing here.
Raise Dynamic range toward 110–120 dB. The slider sets how far below the loudest point is still drawn. At the default 90 dB, faint but genuine high-frequency content can render as black and pass for a cutoff. A band that stays empty at 120 dB is far more convincing.
Hover just below and just above the edge. The readout shows time, frequency and level in dB relative to the loudest point in the file. Note the frequency at the boundary, then compare the levels on either side: a lossy cutoff drops off very steeply over a few hundred hertz.
Scan the whole file, not just the start. A real cutoff is a straight line at the same height from start to finish, through every hi-hat and crash. If the top only looks empty during a soft verse, that's the music, not the format.
Download PNG to compare (optional). Have a FLAC from one source and an MP3 you know is lossy? Save a PNG of each with the same settings. Side by side, matching ceilings are hard to miss.
An audio spectrogram in this tool shows frequencies up to half your device's audio sample rate. On most computers that's about 22 kHz (a 44.1 kHz audio setup) or 24 kHz (48 kHz). The MP3 figures below are the defaults of LAME 3.100, a widely used open-source MP3 encoder, for constant bitrate (CBR) stereo at 44.1 kHz. Use the table as a guide, not a lookup: other encoders, older versions and AAC all differ, and AAC cutoffs vary a lot by encoder and bitrate.
| What you see at the top | What it likely means |
|---|---|
| Hard, flat edge around 16–17 kHz | Typical of a 128 kbps MP3 (LAME: 17 kHz) |
| Hard, flat edge around 18.5–19 kHz | Typical of a 192 kbps MP3 (LAME: 18.6 kHz) |
| Hard, flat edge around 19.5–20 kHz | Typical of a 256 kbps MP3 (LAME: 19.7 kHz) |
| Edge around 20–20.5 kHz | Could be a 320 kbps MP3 (LAME: 20.5 kHz), or a genuine file with little top end. Inconclusive on its own |
| Edge at 15–16 kHz or lower | Low-bitrate lossy encode (LAME cuts 96 kbps at about 15 kHz), or an older encoder; some older AAC files cut this low too |
| Content continues toward ~22 kHz, fading unevenly with the music | Consistent with genuine 44.1 kHz lossless |
| Content continues toward ~24 kHz | Consistent with genuine 48 kHz audio |
| Empty strip only between ~22 and 24 kHz | Normal for a 44.1 kHz file viewed on a 48 kHz device. Not a lossy sign |
Two more clues help on MP3s. First, the band just under the cutoff often looks patchy, with flickering holes and blocks that appear and vanish from moment to moment, rather than the smooth, continuous texture of real cymbals. Above about 16 kHz sits the MP3 format's last scale-factor band (known as sfb21), and encoders like LAME are allowed to code it loosely, so at lower bitrates that top band often comes and goes instead of forming a clean shelf. Second, a lossy edge is straight as a ruler. Real recordings roll off in a ragged way that rises and falls with what's being played.
That last row fools people. When you load a file, the browser converts it to your audio device's sample rate before analysis, so a 44.1 kHz file shown on a 48 kHz device has nothing above 22.05 kHz: a flat, empty strip at the very top. If the edge reads about 22 kHz, the file is behaving exactly as a CD-quality file should.
A spectrogram analyzer is very good at catching an obvious transcode, and much weaker at the edge cases. Before you call a file fake, check whether one of these applies:
Think of it as a test with known false positives. A clean hard ceiling at 16 kHz on a modern pop track is about as close to a verdict as you'll get, while an ambiguous edge near 20 kHz means "can't tell."
The missing frequencies are gone for good. No converter or "restorer" can recover what the encoder discarded, and re-encoding it to a lossy format only adds another round of loss.
For help choosing a format to keep, MP3 vs WAV vs FLAC compares them directly, and bitrate vs sample rate explains which number actually affects the cutoff you just measured.
If the plot itself still feels unfamiliar, how to read a spectrogram walks through every shape and control, and what a spectrogram is covers the theory. To see why a cutoff at 16 kHz versus 20 kHz matters at all, read about the audible frequency range. You'll find more how-to guides in the hub.
Can a spectrogram prove a file is lossless? No. A spectrogram can show strong evidence that a file was made from a lossy source, usually a hard, flat cutoff well below 20 kHz. It can't prove a file is genuine, because high-bitrate MP3s cut off close to 20 kHz and can look almost identical to real recordings. Treat a clean result as consistent with lossless, and a sharp ceiling around 16 kHz as a likely transcode.
Why does my FLAC stop at 16 kHz? Most often because it was converted from a lossy file, such as a 128 kbps MP3 or a low-bitrate AAC, whose encoder removed everything above roughly 16–17 kHz. Converting to FLAC preserved that missing top end exactly. It can also be a naturally dark recording, like an old tape transfer or a solo voice, so check loud, bright passages: a lossy cutoff stays flat and straight through cymbals and hi-hats.
Does converting MP3 to FLAC improve quality? No. Converting an MP3 to FLAC or WAV stores the decoded MP3 without further loss, but it can't restore the frequencies and detail the MP3 encoder already discarded. The result sounds the same as the MP3 while taking up several times the space. It's only worth doing when software you need won't accept MP3, and the file should not be treated as a true lossless original.
Can I check 24-bit/96 kHz files with this spectrogram? Only partly. Vocal Cut's Spectrogram decodes audio at your device's sample rate, usually 44.1 or 48 kHz, so it shows frequencies up to about 22 or 24 kHz. That's enough to catch a hi-res file made from an MP3, since the lossy cutoff sits below 20 kHz. It can't reveal CD audio upsampled to 96 kHz, because that giveaway lies above 22 kHz.
Can you detect a 320 kbps MP3 with a spectrogram? Sometimes, but not reliably. A 320 kbps MP3 from LAME, a widely used encoder, cuts off at about 20.5 kHz, and many genuine recordings have very little energy that high anyway. Raise the dynamic range, use a large window and look at bright passages: a perfectly straight edge near 20 kHz hints at a lossy source, but a soft, uneven roll-off there is inconclusive. Variable-bitrate V0 files can be harder still.