In this guide
Measure the complete final mix in its delivery channel layout. Use the destination’s loudness and true-peak guidance, then verify the encoded file. A sample-peak ceiling or a short-excerpt LUFS result alone cannot certify an episode for delivery.
Podcast loudness needs two different checks: integrated loudness for the overall program, and peak headroom for the highest excursions. Start with the destination’s current instructions, not a preset copied from music streaming. Measure the complete mix you intend to publish, with its actual mono or stereo layout.
A good episode still needs intelligible, reasonably consistent speech. Hitting a meter target does not repair a missing word, a clipped microphone, a loud music bed, or a guest who disappears beneath the host. Resolve those editing problems before treating one final gain change as the solution.
What does Apple recommend for podcast loudness?
Apple’s current podcast audio requirements recommend overall loudness around −16 dB LKFS, with a ±1 dB tolerance, and a true-peak maximum of −1 dB FS. Apple identifies ITU-R BS.1770-5 as the measurement basis and recommends preconditioning before encoding. This is Apple’s guidance, not a universal target or an endorsement of this website.
The same Apple page separates RSS-feed formats from subscriber-audio uploads in Podcasts Connect. Do not transfer one delivery route’s file requirements to another. Confirm the route, channel layout and accepted format before you make a final copy.
You may see integrated loudness labeled LUFS or LKFS. The EBU’s loudness metadata reference explains that these unit names give the same numerical loudness measurement. That does not make integrated, short-term and momentary measurements interchangeable, or prove that two meters use the same settings and scope.
LUFS, sample peak and true peak answer different questions
| Measurement | Useful question | Important limit |
|---|---|---|
| Integrated loudness, LUFS/LKFS | How does the measured program’s overall loudness compare with the target? | A short selection or a different channel layout is a different measurement |
| Sample peak, dBFS | What is the highest stored or decoded sample level? | It does not measure reconstructed peaks between samples |
| True peak, commonly dBTP | How high does a true-peak meter estimate the reconstructed signal reaches? | The final encoding and its decoding still need checking |
The peak-versus-loudness guide explains why normalizing every clip to the same sample peak can leave very different listening levels. For a short example, our free meter reports integrated LUFS and decoded sample peak. It does not provide true peak, loudness range, platform certification or a full-episode workflow beyond its five-minute and 50 MiB limits.
Why mono and stereo must be measured as delivered
Duplicating one mono signal into two identical channels changes the channel energy included in an integrated loudness calculation. It is not the same operation as keeping one mono channel. Do not convert the channel layout after measurement and assume that the integrated number remains valid.
Our existing original reference WAVs make this visible. The mono file’s samples are identical to each channel of the matching stereo file. Both contain the same three-second synthesized tone. The table below reports FFmpeg’s input measurements, before any normalization; it is not a speech example or a result from the browser meter.
| Reference | Channels | Integrated loudness | True peak |
|---|---|---|---|
| One mono channel | 1 | -15.77 LUFS | -12.04 dBTP |
| Two identical stereo channels | 2 | -12.77 LUFS | -12.04 dBTP |
The measured integrated values differ by about 3 LU while the per-channel peak remains the same. That illustrates channel summation in the measurement, not a universal statement that every mono mix should use a target 3 LU below every stereo mix. Panning, processing, changed content and playback routing can also matter. Follow the delivery guidance and measure the actual file.
The measurement report includes source hashes and the native measurement method. The channel guide provides the original listening examples. These existing teaching files are not equipment-calibration signals, AI separation results or a model of a finished podcast’s dynamics.
When gain alone cannot reach the loudness target
Consider an illustrative mix measuring −22 LUFS with a sample peak of −2 dBFS. Reaching −16 LUFS by gain alone calls for about +6 dB, but a −1 dBFS sample ceiling leaves only about +1 dB of sample headroom. These are arithmetic values, not measurements of the reference tones or a true-peak calculation.
The free short-file loudness normalizer applies linked gain and can stop below the requested LUFS target to respect its fixed sample ceiling. It does not use a compressor or limiter to force the target, and that ceiling does not establish a true-peak maximum. Its output report is a working-file check, not an Apple compliance result.
For a complete episode, inspect the event limiting the gain. A single handling bump, an unusually loud laugh, and consistently uneven speech may need different edits. Use your editor’s appropriate level controls or dynamics processing, listening to complete words and transitions, then measure again. Do not repeatedly boost an already clipped file to chase a number.
A podcast export checklist
- Finish the content and mix. Include the introduction, music, guest sections, transitions and ending. Keep an unchanged working original.
- Confirm the delivery. Write down the destination, format, mono/stereo layout, sample rate, integrated target and true-peak limit.
- Measure the complete intended program. Use a meter suited to the specified method and duration. A quiet thirty-second guest excerpt cannot establish the whole episode’s integrated loudness.
- Adjust deliberately. Resolve intrusive events and large speech-level differences before a final linked gain change. Listen for pumping, lost consonants and abrupt changes in the room tone.
- Encode the delivery copy. Keep the working source. Exporting to a lossy format can change peaks and decoded samples.
- Verify the actual encoded file. Check its integrated loudness, true peak, channels, duration and complete content, then listen to the difficult transitions. Do not substitute the export dialog’s requested values for measurements.
If the issue is music competing with speech, use the music-removal workflow before the final level pass and compare any separation with the original. AI WAV downloads require an account and credits. That service is separate from the free local meters and does not automatically produce a mastered podcast delivery.
For storage and timing decisions, the format guide and sample-rate guide explain working copies and conversions. Keep the delivery check tied to the exported file you will actually publish.