Audio Delivery

How to Set Podcast Loudness and Check for Clipping

Meet the delivery loudness target while checking true peaks and retained audio quality after encoding.

A quiet podcast with a few hard peaks can miss a loudness target and already have little peak headroom. Raising the whole track to the target can make those peaks worse. Inspect the source, measure the edit and check the encoded file before delivering it.

Audacity's Loudness Normalization guide explains targeting a loudness value and how its mono handling can affect readings. Apple Podcasts' audio requirements recommend an overall level around -16 dB LKFS, with a tolerance of 1 dB, and a true peak of at most -1 dB FS. Use the actual recipient's specification and measure according to its convention.

Start with the draft mix

Listen to a loud passage in the unprocessed source with the level set comfortably. Mark microphone overload, clipped lines, and badly imbalanced speakers before adjusting an integrated level. Replace a clipped take when a clean alternate exists. Write down whether your meter treats mono as dual mono, then use the same setting for comparable measurements.

Suppose a full episode measures -20 LUFS and -3 dBTP before an adjustment, and the delivery calls for roughly -16. A four-decibel gain change would put that example peak at about +1 dBTP if nothing else changed. Reduce an excessively loud event with clip gain or use appropriate peak control, then measure again. Listen for a dull or flattened laugh, a pumping background, or a voice changing loudness within one sentence. Repeat the calculation with the integrated loudness and true peak from your draft file.

Measure the file that will ship

Normalize the edited programme as a whole when it has the intended balance. Export with the required channels, sample rate and format. Import or measure the exported file and check integrated loudness and true peak again. Lossy encoding and other processing can change performance at the peaks. Use a true-peak-capable meter when your specification calls for one.

Measure the final program with the same set of segments that will appear in the delivery file. A measurement on the loudest two minutes alone cannot stand in for the episode's integrated loudness. Keep the test range in your delivery record if you also measured a separate trailer.

Listen to the loudest edit and to the quietest section you expect people to understand. If the file's integrated number passes but those lines are masked, fix the mix, then recheck loudness and peaks after encoding.

Decide which file you are measuring

Get the delivery requirement before touching the mix. Identify the intended channel count and codec as well as the loudness and peak targets. An episode, trailer, and version with inserted material can each have a different integrated measurement. Measure the exact program the recipient will hear. If a show uses a loud opening and quieter dialogue, measure the whole edit and listen to both. One number does not say whether the voices and music are in balance.

Use the same version of the meter and convention for comparable checks. If you change a mono file into two identical channels, measure that new file with settings appropriate to the requirement. Some meters can account for dual-mono playback separately. Record the channel setting alongside the result so another editor can reproduce it. For more on what is actually in two channels, use the spoken-audio channel guide.

Deal with the event that uses all the headroom

In the example, the mix is four decibels below the intended integrated level but its largest peak is already near the limit. Locate the peak first. It may be a short laugh, hit or music edit that can be corrected individually. Review some dialogue around it and lower the loud event enough to leave room, if that keeps the intended performance. After the edit, measure the entire episode again because the change can affect the integrated result too.

If the peaks are spread throughout a section, test a moderate peak-control approach and compare it with the original at a similar listening level. Keep speech clear and pay attention to the room changing volume after each high point. A very fast reduction may sound different on a laugh, a breath, and a musical transient. Preserve the clean edit in case you need to reach a different recipient specification. If the source take itself is clipped, review the input gain check and use another take if you have it.

Use the loudness and peak measurements together. For an illustration, suppose a revised episode measures -16.4 LKFS with a true peak of -1.7 dB relative to full scale. That would be inside the cited Apple recommendation for both figures. If the encoded version later measures a true peak of -0.8, it is above the recommended peak limit despite the earlier pass. Reduce the offending peak or leave additional margin at the mix stage, then measure the actual encoded file again.

Ask a second listener to check one transition

Play a dialog passage that is meant to be heard over music, followed by the episode's loudest planned cue. Use a phone or headphones at a comfortable setting with the delivery file. Ask whether the listener can understand a quiet name and whether the transition makes them reach for the volume. This can expose a mix decision that is hard to spot in the integrated result. Keep the delivery targets while adjusting the level relationship that failed the test.

Log the exact file name, codec, channels, meter setting and final loudness and true peak. If another editor changes the ending, measure the new export instead of reusing the previous figures. For a later correction, compare the loud event and the full-program numbers under the same settings before replacing the deliverable. A short record also tells the receiver what changed if they hear a different level after the replacement.

If two hosts record different levels across the episode, use the edit to balance their voices before making the final measurement. Compare a passage where one interrupts the other and a section where they speak separately. A normalisation pass over the combined file changes both by the same amount. It cannot repair one speaker repeatedly masking the other. Use individual track or clip adjustments for that difference.

Sources