Loudness normalization versus turning up the volume

Choose the target for the delivery

Apple Podcasts asks for a loudness of about -16 LKFS, within one decibel either way, with peaks no higher than -1 dB.

That target applies to a particular delivery context. Do not treat it as a requirement for every video, music master or broadcast programme. Read the destination’s current guidance and keep the goal in your delivery notes.

The normalizer reports measurements from the source and output. Use those measurements to check the change actually made. A target chosen in a form is an instruction, not evidence that a complicated source reached it exactly.

If the source is silence or has no readable sound, there is no useful speech level to normalize. If the source is clipped, lowering its peaks later does not reverse the distortion already recorded.

Normalize a copy and compare it fairly

A dynamic compressor is a separate tool. It can reduce the difference between loud and quiet parts, but its settings can also flatten emphasis or make background noise more audible. Use it for a diagnosed range problem, not as a compulsory step on every recording.

1. Use the best available audio source and keep it untreated. Listen for clipping, dropout and large differences between speakers before choosing a target.
2. Open the normalizer and select the destination-appropriate target. Run the export and read the measured loudness and peak information.
3. Download the result and listen at a comfortable level. Compare with the source without mistaking a simple level increase for better intelligibility.
4. Check a quiet speaker, a loud laugh and the beginning and end. Overall loudness can look acceptable while a particular passage remains difficult.
5. If the speakers differ greatly, address their balance rather than repeatedly increasing the whole file. Recheck the final combined programme after edits.

Inspect more than the loudness readout

Very short files and mostly silent material can produce measurements that need interpretation. Do not hide a missing measurement behind a confident label. If a number is unavailable, treat it as unavailable and inspect the source.

• Check the measured result rather than only the requested target.
• Listen for new pumping, flattened expression or raised noise during pauses.
• Check true peaks after the final encoding, because encoding can change the waveform.
• Confirm the output sample rate, channels and file type meet the recipient’s requirements.
• Listen to the complete final file after joining, trimming or adding music. Measurements from an earlier intermediate do not approve the finished programme.

Keep the finished mix as the unit of review

A podcast introduction, conversation and ending should be reviewed as a complete programme. Normalizing every short segment independently can change the balance between them.

Keep music beneath speech intentionally. An acceptable overall loudness does not prove that the voice is understandable over a busy soundtrack. Check the passages where the guest speaks most quietly.

Export the final delivery from a retained high-quality source. Save the chosen target and observed result with the version so a later revision can be checked against the same policy. Avoid describing normalization as “studio quality”; state what was measured and what still needs listening review.

Questions about this workflow