Push a file louder without letting anything past a fixed ceiling. Input gain lifts the whole signal, and the limiter turns down only the moments that would have crossed the ceiling.
Lookahead
The limiter delays the audio by 5 ms and analyses the undelayed signal, so a peak is known 5 ms before it is heard. The gain reduction is fully in place by the time that peak reaches the output, which is what stops the distortion a zero-latency limiter produces when it clamps down mid-transient.
That delay is compensated in both directions. The preview keeps the original and processed paths aligned so the bypass button compares the same instant, and the export removes the leading 5 ms so the exported file starts where the source does.
Ceiling
The ceiling is the highest sample value the output is allowed to reach, in dBFS. Nothing crosses it.
Set it to -1 dBFS when the result will be encoded to MP3, AAC, or Opus. Lossy codecs reconstruct a waveform that does not pass exactly through the original samples, and those reconstructed peaks can sit above the ones you limited. A little headroom absorbs that.
For WAV or FLAC that stays as it is, -0.1 dBFS is enough.
Input gain and how hard to push
Input gain is what makes the file louder. The ceiling only decides where the increase stops.
At 6 dB of input gain on a mix that already peaks near full scale, the limiter is working on almost every transient. The result is louder and flatter. At 2 to 3 dB it catches the occasional peak and leaves the rest of the file alone.
Watch the meters while the preview plays. If the level looks pinned at the ceiling for long stretches rather than touching it occasionally, the gain is further up than the material can take.
Release
Release is how quickly the gain comes back after a peak passes. Short releases recover the level fast and hold the average up, but they follow the waveform closely enough on bass-heavy material to distort it. Longer releases stay cleaner and let a loud moment shade the level for a while after it.
150 ms suits most mixed material. Speech tolerates shorter settings than music.
Where a limiter is the wrong tool
A limiter fixes peaks. It cannot make an uneven recording consistent, because it does nothing at all until the signal reaches the ceiling.
If the problem is that some phrases are much louder than others, compress first, then limit what is left. Reaching for input gain alone on that kind of file turns the loud phrases into a flat band at the ceiling while the quiet ones stay where they were.