Normalization
Turning a whole file up or down by one fixed amount to hit a target level, so different clips feel equally loud.
See it
What it is
Normalization scales a whole file up or down by one fixed amount until it hits a target. Nothing inside the file changes relative to anything else: the quiet bits and the loud bits move together, like turning one knob. That is what separates it from compression, which reshapes the distances between loud and quiet.
Two targets, and picking the wrong one is the usual mistake. Peak normalization raises the file until its single loudest sample lands at, say, -1 dBFS. Fast, but useless for matching perceived volume: one stray door slam can hold a whole quiet interview down. Loudness normalization targets an average measured in LUFS (typically -16 LUFS for podcasts, -14 for streaming music), which is what actually makes two files feel equally loud.
Gotcha: normalizing does not clean anything up. Raise a hissy recording and you raise the hiss by exactly the same amount, since the signal-to-noise ratio is unchanged. And loudness normalization can ask for more gain than the headroom allows, so it needs a true peak limiter behind it or the peaks clip on the way to target.
Ask AI for it
Loudness-normalize these audio files to -16 LUFS integrated for podcast delivery, with a true peak ceiling of -1 dBTP. Use loudness normalization, not peak normalization, so perceived volume matches across files, apply one fixed gain change per file rather than compression, and add a transparent limiter after the gain so nothing overshoots the ceiling.