G-XR8P2XJ088

Podcast Audio: Check True Peaks After the Export

A podcast editor levels the conversation, checks the intro music and exports an episode. Nothing on the editing meter turned red. But when the host listens to the delivery file, a sharp laugh near the end sounds rough. Turning down the playback volume makes it quieter without making it clean. The editor now has two files to compare: the master that sounded fine and the encoded copy that revealed the problem.

That example is hypothetical, and a rough laugh could have several causes. What matters is the order of the investigation. Check where the sound first changes before applying another effect to the whole episode. A peak reading from the editing session tells you something about that session. A reading from the delivery file tells you something about the object you intend to publish.

A sample peak and a true peak can differ

Digital audio stores a sequence of measurements called samples. A sample peak meter reports the highest stored level it sees. When the signal is reconstructed for playback, its highest point can fall between those samples. True peak metering estimates that reconstructed level. The November 2023 ITU recommendation on loudness and true peaks describes a measurement method that uses upsampling to estimate what happens between the stored points. Think of it as looking more closely between the marks on a ruler.

That is a different question from how loud the whole episode sounds. Integrated loudness describes the programme across time. True peak concerns its highest short excursion. A quiet interview with an abrupt handclap can have a troublesome peak without being loud overall. A heavily compressed monologue can feel loud even when its highest peaks are held within a ceiling. If the problem is one laugh, making all the dialogue quieter may leave the listener with a new problem: having to turn it up again.

Give the encoder some room and check what comes out

Apple’s podcast audio guidance recommends an overall level around −16 LKFS within one decibel and a true peak no higher than −1 dBFS, with those levels set before encoding. That is a recommendation from one destination, rather than a rule for every host or delivery format. Check the guidance for the file you are preparing. Apple also warns that compression can clip a signal if it has too little peak headroom. This is why a measured master and a listening check of the encoded file complement each other.

Save the lossless master. Make the exact delivery file your host asks for, then open that file in a tool that can measure true peak and integrated loudness. Listen to the loudest moments identified by the meter. If the delivery copy shows a higher peak, compare it with the same moment in the master before deciding what to change. You may need more headroom before encoding. If both files already sound rough at the same point, go back to the unprocessed recording and check whether the sound was distorted during recording or editing.

Pay particular attention to moments when sources overlap. The music may sit comfortably below the host on its own, but a laugh over a drum hit is a different signal from either source separately. If one transition causes the problem, adjusting that transition may preserve the rest of the episode better than pressing every voice against a lower limiter ceiling. Make a fresh export from the master after any change and measure it again.

Make the handover test repeatable

  • Keep a named master and delivery copy. Record the export format and settings. Check the file with the name you will hand over rather than the timeline preview or a previous version.
  • Measure two things. Note integrated loudness and true peak separately. If the peak check points to one problem, listen there before deciding how to fix it.
  • Compare at similar playback loudness. If you listen to the versions at noticeably different levels, a louder file can distract you from the distortion you are trying to identify.
  • Listen to a published preview when available. Check the problem moment on a device listeners are likely to use. If the platform creates a different copy, treat that as one more version to assess.

This check can be small enough to fit a weekly production routine. For the next episode, keep a short note with the version, the reading and one or two timestamps. If a listener reports a problem later, the note tells you what to compare. MaryChuks.com’s guide to judging the sound after compression addresses a related point: the decoded result deserves its own listening pass. Here the decision is whether the final file preserves clean speech and leaves enough room for its highest moments. The meter narrows the search; listening tells you whether the problem matters in the episode.

If you outsource the mix, agree what counts as the final check. A producer might deliver a clean master while someone else makes the compressed file. Name the person who listens to that last copy and give them the original version and the export settings. That small handover can save both people from diagnosing the wrong file.

Save the readings alongside the corresponding versions, too. If the export is replaced at the last minute, the previous numbers no longer describe it. Even a small change to a music cue can alter which moment has the highest peak. Check the whole new file with the meter, then return to the two or three passages that need a human ear.

Sources

Leave a Reply

Discover more from Marychuks.com AI, Psychology, Business & CreativeVerse

Subscribe now to keep reading and get access to the full archive.

Continue reading