Audio analysis · Free

Fake FLAC Checker

Find out whether a “lossless” FLAC, WAV, AIFF, or ALAC file was really made from an MP3 or AAC, upsampled from CD quality, or padded from 16 to 24 bits. Drop a whole album and get a verdict for every track, with the evidence: its spectrogram, the frequency cutoff, and the bits it actually uses.

Runs in your browser · Files are analysed here and never uploaded
Report a problem

Fake FLAC checker workspace

Check files

Is your lossless audio really lossless?

Drop FLAC, WAV, AIFF, or ALAC files, or a whole album. They are analysed one at a time, in the order you add them.

Drop audio files here

FLAC, WAV, AIFF, and ALAC are checked for transcodes, upsampling, and padding. MP3, AAC, Ogg, and Opus are shown as lossy files.

50 files · 1 GB each · 6 GB per batch · no upload

Files are read and analysed in this tab. Nothing is uploaded, stored, or sent anywhere.

What this browser can open

  • WAV, RF64, and AIFF: read directly, in every browser.
  • FLAC: decoded by current Chrome, Edge, Firefox, and Safari.
  • ALAC in M4A: Safari only.
  • MP3, AAC, Ogg Vorbis, and Opus: accepted and labelled as lossy files, where the browser can decode them.

Decoded audio is limited to 1 GB on a computer and 400 MB on a phone or tablet, so very long hi-res tracks may be refused.

Results

Batch results

One row per file, in the order added. Select a file to see its spectrogram and the evidence behind its verdict.

0 of 0 checked

Files you add are listed here with their verdicts. Drop a few tracks above to start.

Selected file

Spectrogram and evidence

Add files above. The first one to finish opens here with its spectrogram, averaged spectrum, and evidence.

Honest limits

What a spectrum can show, and what it cannot.

Evidence, not proof

A spectrogram shows what is in the file, not where it came from. A sharp cutoff is strong evidence of a lossy source; no cutoff is only the absence of that evidence.

Some masters are band-limited

Old recordings, tape, FM and broadcast sources, and masters low-passed on purpose can show a genuine cutoff. Compare other tracks from the same release.

A lossy file can be dressed up

Noise or a synthesised top octave added above an MP3's cutoff makes it look full-band. A pass means no transcode was found, not that one is impossible.

Quiet, short, and narrow files

Speech, instruments with little treble, files under 3 seconds, and very quiet tracks leave too little above the noise floor to judge, and are called inconclusive.

Decoding differs by browser

WAV and AIFF are read directly everywhere. FLAC needs a current browser, and ALAC opens only in Safari. A file decoded at the browser's own rate skips the upsampling and bit-depth checks.

Memory sets the length limit

Decoded audio is held as 32-bit floats: up to 1 GB on a computer (47 minutes of 44.1 kHz stereo) and 400 MB on a phone or tablet. Longer files are refused rather than crashing the tab.

FFT 4,096 to 32,768 points · Blackman window · up to 6,000 frames · loud frames averaged · cliff of 18 dB or more · nothing uploaded

How it works

A spectrum of the whole file, read the way an engineer reads it.

Each file is read in this tab. Its header is read first, for the real sample rate and bit depth, and the audio is then decoded at that rate: WAV and AIFF are read directly, other formats go through the browser's own decoder at the file's rate, so nothing above 22 kHz is lost to resampling. Up to 6,000 frames of 4,096 to 32,768 samples are windowed and transformed, the frames from the loud passages are averaged, and the page looks for the frequency where that average falls off a cliff and stays down. Lossy encoders leave exactly that mark: a steep wall at a predictable frequency. The page also checks whether a high sample rate holds anything above the old one, and whether a 24-bit file uses more than 16 bits.

  1. 01

    Drop your files

    Drop a FLAC, WAV, AIFF, or ALAC file, or a whole album of up to 50. Each one is read in this tab, one at a time, and nothing is uploaded. MP3, AAC, and Ogg files are accepted too, and labelled as lossy.

  2. 02

    Read the verdict

    Each row gets a verdict with its confidence: consistent with lossless, likely transcoded from a lossy source, likely upsampled, padded bit depth, or inconclusive. Select a row for its spectrogram, averaged spectrum, and the evidence behind the call.

  3. 03

    Keep the evidence

    Save the spectrogram as a PNG with the file name and verdict, copy a text report of the whole batch, or download it as a CSV to send to a seller or keep with your collection.

Built for collectors

The checks people run by hand, in one pass over an album.

Native sample rate, no resampling

The header is read before decoding, and the file is decoded at its own rate. A 96 kHz file stays at 96 kHz, so an upsampled master cannot hide behind a browser's 48 kHz default.

A cutoff with its steepness

The averaged spectrum of the loud passages is searched for its highest cliff: at least 18 dB down, nothing coming back above it, and steep enough to be an encoder's wall rather than a natural roll-off.

Encoder signatures

Each cutoff is matched to the low-pass figures in the LAME and FFmpeg source code: 15 to 16 kHz for MP3 at 96 to 128 kbps, 18.6 kHz at 192 kbps, 20.5 kHz at 320 kbps, and the AAC equivalents.

Upsampling and padding

A file at 88.2 to 192 kHz is checked for content above the old Nyquist limit, and a 24-bit or float file for samples that all sit on the 16-bit grid.

Whole albums

Check up to 50 files in order, with a verdict, cutoff, and bits used for each track, then copy a report or download a CSV. Each file's memory is released before the next one starts.

Honest about doubt

Quiet tracks, speech, very short files, and cutoffs where MP3 at 320 kbps overlaps a mastering filter are called inconclusive, with the reason, instead of being forced into a verdict.

Lossless questions

Transcodes, cutoffs, hi-res, and the limits of a spectrogram.

What is a fake FLAC?+

A file in a lossless format whose audio came from a lossy one. Someone takes an MP3 or AAC, decodes it, and saves the result as FLAC or WAV. The file is bigger and every bit of it is preserved, but what was preserved is the sound of the MP3: the detail the encoder removed is gone, and no conversion can bring it back. The same happens with hi-res: a CD-quality master converted to 96 kHz or 24 bits gains size and no information.

How does a spectrogram reveal an MP3 source?+

To save bits, lossy encoders drop the highest frequencies first, with a low-pass filter far steeper than anything in nature. Decoded and saved as FLAC, that leaves a sharp horizontal edge in the spectrogram: full detail below one frequency and almost nothing above it, for the whole track. Genuine CD audio normally continues to about 20 to 22 kHz, where the converter's own anti-alias filter stops it.

Which cutoff points to which bitrate?+

From the LAME encoder's own table: about 11 kHz at 64 kbps, 13.5 kHz at 80 kbps, 15.1 and 15.6 kHz at 96 and 112 kbps, 17 kHz at 128 kbps (often with nothing above 16 kHz, where MP3's highest band starts), 18.6 kHz at 192 kbps, 19.4 to 19.7 kHz at 224 to 256 kbps, and 20.5 kHz at 320 kbps. FFmpeg's AAC encoder cuts at 16 kHz at 128 kbps, 18 kHz at 192, and 20 kHz at 256. Other encoders and settings differ, so the page names likely sources rather than one.

Can a genuine lossless file have a cutoff?+

Yes. Some masters are low-passed on purpose, old recordings and cassette or FM sources have little above 15 or 16 kHz, and many resampling and mastering filters stop near 20 kHz. That is why a cutoff above 20.2 kHz is reported as inconclusive, why the page measures the steepness and how many loud frames share the drop, and why each verdict is a likelihood with its evidence. Check other tracks from the same release before deciding.

How do I tell whether a 24-bit or 96 kHz file is really hi-res?+

Two checks. If a file at 88.2 to 192 kHz has nothing above 22 or 24 kHz, it was almost certainly upsampled from 44.1 or 48 kHz, and the page names the likely original rate. If every sample of a 24-bit file sits exactly on the 16-bit grid, its bottom 8 bits are zero and it holds 16-bit audio padded with zeros. Browsers decode 24-bit samples into 32-bit floats without rounding, so this test is exact.

Why is my MP3 or M4A labelled as a lossy file?+

Because MP3, AAC, Vorbis, and Opus are lossy formats, so whether they are lossless has an obvious answer. The page still shows their spectrogram and cutoff, which helps tell a 128 kbps file from a 320 kbps one, or check what a store actually sold you.

Which formats and browsers work?+

WAV, RF64, and AIFF are read directly by the page, so they work in every browser. FLAC is decoded by the browser in current Chrome, Edge, Firefox, and Safari. ALAC in M4A decodes only in Safari. AAC in M4A and MP3 work in all of them, while Ogg Vorbis and Opus work in Chrome, Edge, and Firefox and depend on the version in Safari. When a file cannot be decoded, its row says so: open ALAC in Safari, or convert the file to FLAC.

How large a file can it check?+

Up to 50 files and 6 GB per batch, and 1 GB per file. The real limit is memory: decoding holds the whole track as 32-bit floats, so the page refuses a file whose decoded audio would pass 1 GB on a computer or 400 MB on a phone or tablet. On a computer that is about 47 minutes of 44.1 kHz stereo, 21 minutes at 96 kHz, or 10 minutes at 192 kHz. Files are checked one at a time and released before the next.

Can a lossy file be made to pass?+

Yes, and that is the main limit of any spectral check. Noise, or a synthesised top octave, added above an MP3's cutoff makes the spectrogram look full. Such a top band often has no structure and does not follow the music. Treat a pass as no evidence of a transcode, rather than as proof of a lossless origin.

Does anything get uploaded?+

No. Files are read and analysed in this tab, with the browser's own audio decoder and a Fourier transform that runs in JavaScript. Nothing is sent to a server or saved in the browser. A PNG, report, or CSV exists only when you ask for one, and goes straight to your device.

More focused tools, ready when you are.

Explore the growing collection for calculations, documents, writing, and everyday work.

Browse all tools