Audio Quality Checker — What Is Actually In This File?
Upload a single recording. DiffALL measures integrated loudness in LUFS, finds clipped samples, reads how much of the frequency range is really there, and says whether a stereo file carries two channels or one twice.
Free, no sign-up — your files never leave the comparison.
Drop an audio file here or click to browse
MP3 · WAV · M4A · FLAC · OGG — up to 100 MB
Free, no sign-up — your audio is deleted as soon as the check finishes.
Reading the audio…
What DiffALL measures
Loudness in LUFS
Integrated loudness to ITU-R BS.1770, the measurement every streaming platform normalises to. The filters are derived at your file’s own sample rate rather than assuming 48kHz, which is the usual bug in loudness code.
The bitrate it really carries
A 64kbps file re-exported at 320 still has 64kbps of content in it. The frequency range says so — measured, a laundered file reads content to 10.9kHz while its header claims 320kbps.
Clipping, counted properly
Not just samples at full scale, but the longest flat run of them. One sample touching the ceiling is a peak; a dozen in a row is a waveform with its top cut off. A master limited to -1dB is not called clipped.
Is the stereo real?
Side content against mid, so a file that is stereo in its header and mono in its content is caught — twice the size for no benefit. Correlation cannot see this; measured, real stereo and dual mono differ by 0.001.
It says when it cannot tell
There is no noise check here, and that is deliberate: no measurement separates hiss from quiet content without a reference, and a check that calls a field recording noisy is worse than no check.
Supported formats
Use cases
Before you publish
Check a master against the -14 LUFS every platform normalises to, and catch clipping before the encoder makes it worse.
Is this really the master?
A file handed to you as 320kbps may have been 128kbps earlier in its life. The frequency range shows what is actually in it.
Check a podcast export
Loudness, clipping and whether your chain quietly folded a stereo mix down to one channel duplicated.
Triage a recording session
Find the takes that were recorded too hot before you start editing them.
Frequently asked questions
How do you check the quality of an audio file?
Upload one file. DiffALL samples one-second windows spread across it and measures five things: integrated loudness in LUFS against the streaming target, clipped samples and how long the flat runs are, the highest frequency the file actually carries, bitrate expressed as bits per sample, and — for stereo files — whether the two channels differ at all. Each gets a band from Excellent to Poor, and the weakest is named at the top.
What is LUFS and why -14?
LUFS is loudness as a listener perceives it, defined by ITU-R BS.1770: the signal is filtered to match how hearing weights frequency, then averaged. Spotify, YouTube, Apple Music and Tidal all normalise playback to around -14 LUFS, so a master far above it simply gets turned down — the extra level buys nothing and costs dynamic range. Broadcast uses -23.
Can you tell if my file was once a low-bitrate MP3?
Often, yes, and it is the most useful thing this tool finds. Lossy encoders throw away the top of the frequency range, and re-exporting at a higher bitrate does not put it back. Measured on one source: 320kbps keeps content to 19.9kHz, 128kbps to 16.5kHz, 64kbps to 10.9kHz — and a 64kbps file re-encoded at 320kbps still reads 10.9kHz while its header says 320. What the check cannot do is tell a codec from a deliberate low-pass filter, so it reports what is in the file rather than guessing how it got that way.
Why is there no noise or hiss check?
Because it cannot be done honestly without something to compare against. A noise floor is easy to measure, but nothing separates hiss from quiet content: measured across our test set, music sits at 11.8dB of dynamic range and a field recording at 6.0dB, against 17.4dB for speech ruined by audible noise. Any threshold that catches the ruined file condemns the music. A check that calls good work broken is worse than no check.
How much of my file gets checked?
The free plan reads the first 60 seconds; Pro reads up to 3 minutes. Either way the windows are spread across that span rather than taken from the opening, and the report says how many seconds it sampled.
Is my audio stored on your servers?
No. The file is deleted as soon as the check finishes. Nothing is kept and nothing is inspected.
Ready to compare audio files?
Free, no install, no sign-up required for your first comparison.
Start comparing now ↑