Audio Analyzer

Free for Windows

Is your lossless music really lossless?

Plenty of FLAC and WAV files out there were made from an MP3 or AAC and converted back. They look lossless, take the space of lossless, and sound like the MP3 they came from. Audio Analyzer finds them, and shows you the evidence.

Download for Windows Portable zip and release notes on GitHub
Audio Analyzer checking four songs. The selected one, Track 01, is marked Lossy source with high confidence: its spectrogram stops dead at 16.8 kHz, the cutoff of a 128 kbps MP3.
A FLAC made from a 128 kbps MP3: everything above 16.8 kHz is gone, and the app says so.

The problem

How a fake lossless file is made

Converting a lossy file back to FLAC can't bring back what the encoder threw away. The file is bigger, the label says lossless, and the detail is still missing.

✓ Lossless

The original

A CD or studio master. Every detail kept, up to the top of the audible range.

✗ Lossy

Squeezed to MP3 or AAC

The encoder deletes what it thinks you won't miss: the highest frequencies and many quiet details.

✗ Still lossy

Converted back to FLAC

Same file type as the real thing, many times the size of the MP3, and none of the lost detail.

Genuine lossless

Sound reaches about 21 kHz, and its top edge moves with the music.

Spectrogram of 20 seconds of a CD-quality song: bright, uneven content all the way up to about 21 kilohertz.

The same 20 seconds, via a 128 kbps MP3

Everything above 16.4 kHz is gone: a flat wall, the mark of an encoder.

Spectrogram of the same 20 seconds after conversion to a 128 kilobit MP3 and back: content stops at a flat line at 16.4 kilohertz, with only black above it.

How it checks

Four tests on every file

Each one looks for a different trace that lossy encoders leave behind. Some fakes leave only one.

Brick-wall cutoff

Encoders delete everything above a fixed frequency: about 16 kHz for a 128 kbps MP3, 19 kHz for 192 kbps. Real recordings fade out instead.

CatchesMost MP3, AAC, Opus and Vorbis files

Spectral holes

Encoders blank out narrow high-frequency slices while the slices next to them keep their sound. Natural recordings almost never do that.

CatchesHigh-bitrate files with no visible cutoff

AAC block fingerprint

AAC and Vorbis work in blocks of 1,024 and round many values to exactly zero. Lined up on that grid, the zeros reappear.

Catches256 kbps AAC that looks perfect on a spectrogram

Container honesty

A 96 kHz file with nothing above 22 kHz was upsampled from CD quality. A 24-bit file that uses only 16 bits was padded with zeros.

CatchesFake hi-res

One clear verdict, with a confidence level

Lossless

No sign of a lossy source anywhere in the file.

Lossy source

Made from a lossy file, with a guess at the original bitrate.

Upsampled

Sold as hi-res, but converted up from CD quality.

Inconclusive

Some clues, not enough to accuse. It says which, and why.

Tested, not guessed

Measured on 3,442 files with a known origin

440 known-lossless files (CD rips verified against AccurateRip, the EBU test CD, live soundboard recordings and studio masters) and 3,002 known fakes made from them in 14 different ways.

0

of 280 CD-verified tracks wrongly called lossy

1 in 440

genuine files wrongly accused overall (a soundboard recording with a hardware filter)

3 in 4

fakes caught outright; most of the rest are flagged, not cleared

100%

on your PC: no upload, no AI model, no subscription

Fakes caught outright, by how they were made

Share of 215 test files per format called "Lossy source". Files not caught are mostly marked Inconclusive.

  • AAC 256 kbps93%
  • AAC 128 kbps92%
  • MP3 128 kbps90%
  • Ogg Vorbis q587%
  • MP3 192 kbps84%
  • MP3 256 kbps80%
  • MP3 320 kbps72%
  • MP3 V0 (VBR)44%
  • Opus 128 kbps40%

MP3 V0 and Opus are the hardest: they often keep the full frequency range and leave no holes. When the app can't be sure, it says so rather than calling the file clean.

More than a checker

Everything you need to trust your library

Compare versions

A CD rip, a download and a remaster of the same song, side by side. It recommends the best copy, and Listen for the differences plays the same moments in each so you can hear them.

Scan a whole folder

Check an entire library in one go, with FLAC integrity checks and duplicate finding. Every result is remembered, so a re-scan only looks at what changed.

A written report

Why each file is or isn't lossless, in plain language, with every measurement. Save it as Markdown or plain text.

ReplayGain

Checks your ReplayGain tags and fixes them if you ask. The audio itself is never touched.

Untouched playback

Bit-perfect through WASAPI exclusive mode where your sound card allows it: no effects, no resampling, no Windows mixer.

Keeps itself up to date

Installs for you only, no administrator password. When a new version comes out, the app offers it.

A FLAC made from a 256 kbps AAC: the spectrogram looks complete up to 22 kHz, but the app reports Lossy source because the AAC block fingerprint was found.
A FLAC made from a 256 kbps AAC looks perfect on a spectrogram. The block fingerprint still gives it away.

Honest limits

What no tool can do

  • Prove a file is lossless. "Lossless" means no evidence of a lossy source was found.
  • Settle 20.4 to 21.3 kHz cutoffs. A 320 kbps MP3 and some CD masters look the same there, so those files get Inconclusive, never a false accusation.
  • Judge speech well. Lossy encoders leave their marks in the high frequencies, and speech has little there.

Privacy

Your music stays on your PC

  • Nothing about your music is collected or uploaded.
  • It changes a file only when you ask it to write ReplayGain.
  • Once a day it asks GitHub whether there's a newer version. You can turn that off.
  • Send feedback sends what you write, only when you press Send.

Find out what's really in your library

Free. Windows 10 and 11, 64-bit.

Download for Windows

The app isn't digitally signed yet, so Windows may say "Windows protected your PC" the first time. Click More info, then Run anyway.