Finagle My Song · Blog
The MP3 ghost hiding in your "WAV" file
A WAV file only proves the container is lossless. What actually happened to the audio before it landed in that container is a separate question, and a lot of files that look pristine on paper have already been through an MP3 once.
How a file gets a ghost
It doesn't take much. A sample pack downloaded as MP3 and bounced into a session as WAV. A reference track saved off a streaming rip, dropped into the mix to match against. A rough mix sent for approval as an MP3, then that same file mistakenly becoming the source everyone masters from instead of the real bounce. A vocal recorded through some app that compresses audio internally before it ever reaches your DAW.
In every case, the file you're holding now is a WAV. Full 16 or 24-bit, uncompressed, opens fine, measures fine on a peak meter. But the audio inside it already ran through lossy compression at some point upstream, and that step isn't reversible. Whatever got thrown away is gone permanently — converting back to WAV afterward just stores the damage more expensively.
What actually gets thrown away
MP3 and other lossy codecs save space by cutting content a listener is less likely to notice, and the easiest thing to cut is the very top of the frequency range. At lower bitrates especially, encoders apply a hard low-pass filter, often somewhere in the 15–19kHz region, and simply discard everything above it.
Real recorded and mixed audio doesn't do this. A microphone, a synth, a full mix bus — they all have a natural, gradual high-frequency roll-off with no sharp edge. A lossy-encoded source has a cliff: content, then a hard stop. That cliff is a fingerprint. Our analyzer compares the energy above 15kHz with the energy between 4 and 15kHz, and when there's almost nothing up top (under 2% of the band below it), it flags the track as a suspected lossy source. Even dark, real recordings keep more than that.
Why this matters more once you're mastering
A mix engineer or an AI mastering chain can shape whatever frequency content exists in a file. Neither can add back content that was never captured. If your source already lost everything above 16kHz, no amount of "add air" or high-shelf boosting brings that detail back. It just boosts whatever noise or artifacts are left in the band that survived, which usually makes the file sound harsher, not brighter.
That's the frustrating part for anyone who doesn't know their source is compromised: they hear a master that sounds dull or thin no matter what they try, assume the mastering is the problem, and keep pushing for brightness that was never going to be there.
What to actually do about it
First, check whether a real lossless bounce exists somewhere upstream — the original session export, a WAV from before it got shared as an MP3, the actual multitrack instead of a rough mix. If it exists, use that as your master source instead.
If it doesn't, if the only version left really is this one, stop chasing top end that isn't there. Accept the ceiling and put your mastering decisions where they'll actually matter: loudness, dynamics, and tonal balance in the range that's intact. And going forward, treat MP3 as a delivery format only, never a working format — anything you might master, remaster, or reference later should get bounced and stored as WAV or FLAC from the start.
Finagle My Song's free analysis flags a suspected lossy source automatically, before you spend time mastering a ceiling that was set months ago by someone else's export settings.