Why your audio clips after mastering

This one is genuinely unfair. You did everything right: mastered your episode, set a limiter with a -1 dB ceiling, watched the meter never touch red, exported. Then you play the published version and the loudest consonants have a hard, brittle edge that was not in your export. Where did it come from?

Your meter measures dots. Your listener hears the curve.

Digital audio stores samples: tens of thousands of measurement points per second. Playback does not output those dots directly. The converter reconstructs the smooth analog curve that passes through them, and here is the trap: the curve can swing higher between two samples than either sample. A peak meter only reads the dots, so it reports -1 dB while the reconstructed waveform briefly exceeds 0, which is physically more level than the output stage can produce. The result is clipping your export never contained.

These are inter-sample peaks, and the measurement that catches them is called true peak, written dBTP. The standard method (ITU-R BS.1770) oversamples the signal 4x, roughly guessing the curve between the dots, and reads the maximum of that instead.

Lossy encoding makes it worse

Almost nobody hears your WAV. They hear an MP3, AAC, or Opus version created by the platform, and lossy codecs do not reproduce waveforms exactly. Encoding adds small overshoots of its own, commonly a few tenths of a dB and sometimes more on dense material. A master that sits at exactly 0 dBFS sample peak can come out of an MP3 encoder with true peaks well above that.

This is why broadcast standards and platform guidelines converge on a ceiling of -1 dBTP, and why the number is stated in dBTP rather than plain dB. The spare decibel is not caution theater. It is budget for the encoder.

Why your limiter did not save you

Most limiters, including the stock one in nearly every DAW, are sample-peak limiters. They guarantee no sample exceeds the ceiling and say nothing about the curve between samples. Transients (drum hits, plosive consonants, mouth clicks) are exactly where the curve moves fastest between dots, which is why the damage always concentrates there.

True-peak limiters exist. They oversample internally and limit the reconstructed curve instead. They are also slightly slower and can sound slightly less transparent at extreme settings, which is a fair trade for output that survives encoding.

How to actually fix it

• Measure with a true-peak meter, not a sample meter. If your export reads above -1 dBTP, it is at risk.
• If your limiter is sample-peak only, set its ceiling lower (-1.5 to -2 dB) to leave room for what it cannot see.
• Better, verify after the fact: measure the final file's true peak and trim if needed. A 0.5 dB trim is inaudible; encoder clipping is not.
• Do not chase loudness past your platform's normalization target. Extra limiting buys you nothing after normalization, and it manufactures exactly these peaks. Numbers in our loudness guide.

One honest caveat: if the distortion was recorded at the source, meaning the microphone or preamp clipped, none of this applies. That damage is baked into the waveform itself, and no ceiling fixes it. True-peak discipline protects clean recordings from being damaged in delivery. It cannot resurrect ones that arrived damaged.

FixAudio handles this chain automatically: masters to target loudness, then measures the true peak of the finished file with 4x oversampling and trims to keep it at or under -1 dBTP. We added the verification step after watching a limiter alone let +2.4 dBTP peaks through on our own test corpus. Meters are cheap and honest. Measure the file you actually publish, not the one your limiter promised you.