Free · no login · nothing uploaded
Is that file what it says it is?
The frequency cliff that gives away an MP3 wearing a FLAC extension, whether the peaks are already squared off, and the one no other checker measures — what happens to your low end when a club system sums it to mono. All of it on your machine.
Drop a file you are not sure about
WAV, AIFF, FLAC, MP3 or M4A. Select a folder’s worth and they queue. Nothing is uploaded — this page makes no network request at all.
The check nobody else runs
A club system will sum your bass to mono.
Every quality checker on the web reads the frequency cliff, and it is worth reading. But the failure that actually costs a producer a dancefloor is invisible to it, invisible on headphones, and completely measurable.
Below roughly 120 Hz a large sound system has no stereo image and cannot have one. The wavelength at 60 Hz is close to six metres, which is wider than the space between anyone and the subs, so a subwoofer array is fed a mono sum by design. Whatever your left and right channels are doing to each other down there decides how much of your low end arrives in the room.
If the two are in phase they add and nothing is lost. If they have been widened — by a stereo imager across the whole mix, by a sub layered with a detuned copy of itself, by a sample that was already wide — they partly cancel, and the energy goes nowhere. On headphones each ear hears its own channel, so nothing cancels and the track sounds enormous. The first time you find out is on the system.
So this tool measures the correlation between the channels inside each band and, more usefully, the decibels each band loses when the two are added together. Anything worse than -3 dB in the sub row is half your low end not reaching the floor, and it is a mix decision rather than a file problem — which means it is fixable, and only if you know.
The cutoff detector, calibrated against known bitrates
The whole lossless-or-not verdict rests on one claim: that the corner where the energy stops moves with the encoder’s bitrate and can be read back out. That is testable, so we tested it — on 49 records encoded at bitrates we chose, which means the answer was known before the detector saw the file.
11.2 kHz
At 64 kbps
13.3 kHz
At 80 kbps
16.7 kHz
At 96 kbps
100%
Separated on one threshold
Method: the same 49 records encoded with LAME at each bitrate and measured with the detector this page runs. The sources are 96 kbps store previews, so the test only runs downwards — re-encoding upward cannot restore bandwidth that is already gone, which is the same fact the tool exists to exploit. What this shows is that the corner tracks the encoder, and by how much. It is not a claim about the absolute threshold for 320 kbps material, which we have no lossless corpus to test. Measured 2026-08-11.
How it works
What the checker actually reads
Three measurements, and each one answers a question you would otherwise have to guess at.
Where the audio stops, and why that is a fingerprint
Every lossy encoder throws away the top of the spectrum before it does anything else, because that is where the ear notices least and the data costs most. The frequency it chooses to stop at is a property of the bitrate: roughly 16 kHz at 128 kbps, 19 kHz at 256, 20 kHz and up at 320 or lossless. That corner survives everything you do to the file afterwards. Re-encode a 128 kbps MP3 to 24-bit FLAC and you get a large file with a cliff at 16 kHz, because no encoder can put back what the first one deleted. The corner is measured relative to the track's own mid-band level, so a deliberately dark master is not mistaken for a lossy one.
What the low end does when it is summed
A club sound system does not reproduce a stereo image below its crossover — the subs are a mono array, normally fed everything under about 120 Hz. So whatever your two channels do to each other down there decides how much bass reaches the room. If the sub is wide, or worse out of phase, the two channels cancel when they are summed and the low end simply is not there on a big rig, while sounding enormous on headphones. This is measured directly: the correlation between the channels inside each band, and the decibels the band loses when they become one.
Whether it is already clipped, and whether a player will take it
Clipping is counted as flat tops rather than as loud samples. One sample pinned at full scale is a peak that landed on the ceiling and is inaudible; three or more in a row is a limiter that ran out of range and squared the waveform off, which is distortion baked into the file. Alongside that, the true peak is read at four times the sample rate, because the waveform between the stored samples routinely goes higher than any of them. And the container is checked against what a CDJ will actually read off a stick, which is a different list from what your laptop plays.
Reference
Reading a cutoff
Where each bitrate stops, and what it means for a file you were about to put on a stick.
| Format | Energy stops at | What it means for a set |
|---|---|---|
| Lossless (WAV, AIFF, FLAC) | to Nyquist — 22 kHz at 44.1 kHz | Nothing was removed. This is the only case where the top octave is genuinely there. |
| MP3 320 kbps | 20 kHz and up | Effectively transparent on a system. What a DJ store sells and what belongs on a USB stick. |
| MP3 256 / AAC 256 | 19 to 20 kHz | Fine to play. Indistinguishable from 320 on almost anything. |
| MP3 192 | 18 to 19 kHz | Acceptable at home, audibly softer at the top on a big rig. |
| MP3 / AAC 128 | 15 to 16 kHz | Thin on a system, and the first thing a sound engineer will notice. Replace it. |
| Under 128 | below 15 kHz | A rip, a stream capture or a YouTube download. Do not play this out. |
These are the corners LAME and the common AAC encoders pick by default, and an encoder can be told to do otherwise. Read a cutoff as strong evidence rather than as proof — which is why the tool marks a lossless file suspect below 19 kHz and only calls it outright below 17.
Frequently asked
Common questions
Does my file get uploaded?+
No, and this page is easy to verify: open the network panel before you drop a file and watch it stay empty. This tool does not ask our catalog anything either — every verdict it prints is computed from your own samples, in your own tab, by JavaScript that is already in the page. There is no upload endpoint behind it.
How can you tell a FLAC was made from an MP3?+
Because the MP3's fingerprint is still in it. Every lossy encoder low-passes the signal before it quantises — it deletes the top of the spectrum, permanently — and the frequency it stops at is set by the bitrate. Converting the result to FLAC stores that band-limited audio losslessly, which means it stores the cliff too. So a file whose container says 24-bit 96 kHz and whose audio stops dead at 16 kHz was a 128 kbps MP3 an hour ago, and nothing about the container changes that. The tool shows you the spectrum with the corner marked, so you can see it rather than take our word for it.
Could a real lossless file have a low cutoff?+
Yes, and it is the one thing that makes this a judgement rather than a test. A deliberately dark master, an old recording, or anything sourced from tape can genuinely stop early. That is why the threshold for calling a lossless file suspect is 19 kHz rather than 20 — between the two sits an honest dark master, and it gets a "worth knowing" rather than an accusation. Below about 17 kHz there is no ambiguity left: no mastering choice produces a vertical cliff at exactly the frequency an encoder would have picked.
How reliable is the cutoff reading itself?+
Measured, on files whose bitrate we knew because we made them. Encoding the same 49 records at three known bitrates, the detector reads a median corner of 11.2 kHz at 64 kbps, 13.3 kHz at 80, and 16.7 kHz at 96 — separating the highest and lowest of those correctly 100% of the time on a single threshold. The corner moves where the encoder moves it, which is the property the whole method rests on.
What does "the bass will not survive a club system" actually mean?+
It means the two channels are cancelling each other below about 120 Hz. A large sound system feeds its subwoofers a mono sum, because a subwoofer array physically cannot reproduce a stereo image at those wavelengths — the room is smaller than the wave. So if your kick and your sub are wide, or a stereo widener has been applied across the whole mix, the left and right low end partly cancel when they are added together and the energy disappears. On headphones each ear hears its own channel and nothing cancels, which is why this is a mistake that survives all the way to the club. The fix is to make everything below the crossover mono in the mix.
Which formats will a CDJ actually play?+
MP3, AAC/M4A, WAV and AIFF on everything. FLAC on a CDJ-2000NXS2 and newer only, so if you do not know what is in the booth, AIFF is the safe version of the same audio. Ogg and Opus on nothing — no Pioneer player reads them and Rekordbox will not import them either. The tool says which of those you are holding and what to convert it to.
What counts as clipping?+
Runs of consecutive samples pinned at full scale. A single one is a peak that happened to land on the ceiling and you will never hear it; three or more in a row means the waveform has been squared off, which is distortion recorded into the file and cannot be undone by turning it down. Separately, the true peak is measured at four times the sample rate, because a file can have no pinned samples at all and still go over full scale in between them — which is what a converter and a lossy decoder both reconstruct.
Is a 320 kbps MP3 good enough to play out?+
Yes. The difference between 320 kbps and lossless is not something that survives a PA, a room and a crowd, and every DJ store sells 320 as a standard tier for that reason. The bitrates worth caring about are the ones below it: 128 kbps is audibly thin at the top on a big rig, and anything under that is a rip. The reason to buy lossless is that it is the format you keep — you can make a 320 from a WAV forever, and you can never go the other way.
Can I check a whole folder at once?+
Select as many files as you like and they queue, one at a time, with every result staying on screen and a CSV export once there is more than one. For a whole collection rather than a folder, the library health check reads a Rekordbox export and measures the lot at once — also without uploading anything.
Explore
LUFS meter · Key & BPM finder · Library health check · Harmonic path finder · Lossless · Lossy compression · Bitrate · FLAC · WAV and AIFF · Mono sum and phase · All tools
Build a set from tracks that fit
DJ terms on this page: Lossless · Lossy compression · Bit rate · FLAC · WAV and AIFF · Mono sum and phase · True peak · Headroom · Limiter · Waveform · Full glossary