Audio Comparison

How the Audio Comparison works

Audio comparison: what it does and why

Comparing two pieces of audio honestly is harder than it sounds. Louder almost always sounds better, so a comparison where one file is a decibel hotter is decided before you have listened; and switching between two players by hand introduces enough delay to forget what the first one sounded like. This tool renders the comparison into a single file with the level bias removed, so what you hear is the difference and nothing else.

A and B. The first file in the list is A and the second is B; any further files are ignored, and the stats tell you so. For each the tool reports the duration, channel count and sample rate, the peak and RMS level, and the integrated loudness in LUFS measured with the same algorithm as the Loudness Meter, followed by how many decibels louder or quieter B is than A. That last number alone answers a lot of questions — how much a limiter added, whether an encoder changed the level, whether two releases of the same album were mastered differently.

Loudness matching. With Match loudness on, B is trimmed by the LUFS difference so that both play at the same perceived loudness in every render mode. B is never boosted by more than 12 dB, so a much quieter B plays quieter than A. The stats show the trim applied. Turn it off when the level difference is itself the thing you want to hear, or when comparing a file against a silent or nearly silent one where the match would be meaningless.

Alternating. The default render plays A for five seconds, then B for five, then A again, to the end of the longer file. Two-second switching is more revealing for small differences in top end or transients, because the memory of the previous segment is still fresh; five seconds suits judging tone or space. Listen without looking at the clock and try to say which segment is which — if you cannot, the difference is not one that matters.

A only / B only. These render one file at the matched level, which is useful for making a loudness-matched copy of B to load into another player or to hand to someone else for a blind test.

Difference. This is the null test. B is inverted and added to A, so anything the two files share cancels and only what differs remains. Compare a WAV with its MP3 encode and you hear the encoder's discarded material — usually faint high-frequency chatter; compare a mix before and after a compressor and you hear exactly what the compressor changed. A result that is completely silent means the files are identical; a result that sounds like the full mix means they are not aligned in time or one is polarity-inverted, and the comparison is not valid. The tool warns when the lengths differ by more than 50 ms, which is the commonest reason a null test fails, since encoders often add a few milliseconds of padding.

Format handling. The files are rendered at A's sample rate with as many channels as the larger of the two, so a mono and a stereo file, or a 44.1 kHz and a 48 kHz file, can still be compared. The output is a WAV named after the source.

Related tools