Start with the requested gain
The first gain estimate is the target integrated loudness minus the measured source loudness. Applying a constant gain changes peak level by the same number of decibels and retains the relative dynamics. If the requested boost would exceed the chosen true-peak ceiling, the preserve-dynamics mode applies less gain.
That result can be quieter than the target. The tool reports the difference explicitly rather than hiding clipping in the export. For an already loud source, reducing gain is often sufficient. Silence or material below the loudness gate has no meaningful integrated value to normalize and is refused in processing mode.
Understand what limiting changes
Linked limiting uses a shared envelope for both channels so a transient does not independently move the stereo image. Offline pre-attenuation and a release envelope reduce excessive peaks before the final file is written. This is dynamic processing; it can change attack, sustain and perceived punch.
A sample-limited waveform can still produce intersample peaks. The workbench therefore measures true peak again and applies a final uniform reduction when necessary. A larger reduction can move the achieved integrated level away from the requested target. The report keeps both the limiter reduction and the final correction visible.
Check the actual output file
The finished result is encoded as 24-bit PCM WAV, decoded again, and remeasured. A file with remaining full-scale samples or a measured peak above the chosen tolerance is not offered as a successful export. This verifies the downloadable artifact rather than only the floating-point working buffer.
Listen to transient-rich passages and quiet material near louder events. Compare before and after without automatically choosing the louder version. A downstream AAC or MP3 encoder may create new peaks, so this PCM check is not a guarantee for an untested codec. Keep the original source and the before/after report.