Speech workflows
Set consistent levels before you publish
Raise a quiet recording, even out phrases, hold the peaks, and confirm the result on the meter.
Hear and see the problem
Play a loud passage and a quiet one while watching the meter in the transport bar. Peaks that never pass −12 dBFS mean the whole recording is quiet; peaks that hit 0 mean it is already clipping and needs its gain reduced rather than raised. A large gap between the RMS reading in loud and quiet phrases means the dynamics need evening out.
Fix noise first, because the steps below raise everything, including background noise.
Compress, then set the ceiling
- Select all with Cmd/Ctrl+A and choose Process → Compressor…. Keep the default 4:1 ratio for speech, set the threshold so only the louder phrases are reduced, and add enough makeup gain to bring the quiet phrases up. Preview and apply.
- Choose Process → Limiter…, keep the −1 dBFS ceiling, and lower the threshold until the meter shows quiet phrases close to the loud ones without audible pumping.
- If the result is still too quiet or slightly too loud overall, finish with Process → Normalize… to a −1 dBFS peak.
Each step is one undo, so compare with Cmd/Ctrl+Z and Cmd/Ctrl+Shift+Z at the same listening volume.
Check the export
Play the whole recording once more and watch the peak reading; it should stay at or below the ceiling. Export as MP3 or WAV and listen to the export as well, because MP3 encoding can raise peaks slightly. Save the project so you can revisit the settings, and use Repeat Last Effect on the next episode to apply the same limiter settings.