Every time someone streams your podcast through Apple Podcasts or Spotify, the player runs a normalization pass on the audio before it reaches their earphones. The target for both platforms is -16 LUFS integrated loudness. If your master is louder than that, it gets turned down. If it's quieter, it gets turned up. The normalization is not optional and it's not under your control — which means the only thing you can control is what you hand to the platform.
Understanding what LUFS means, how the measurement works, and what it means for your master bus chain is the difference between a show that sounds consistent across every player and episode and one that sounds fine in your headphones but weak, clipped, or uneven when listeners hear it in their cars.
What LUFS actually measures
LUFS stands for Loudness Units relative to Full Scale. It's the unit defined in EBU R128, the European Broadcasting Union's loudness normalization standard published in 2010 and subsequently adopted in broadcast and streaming audio worldwide. The companion standard from the Audio Engineering Society — AES TD1008 — provides recommended practices specifically for streaming audio delivery, and both converge on the same underlying measurement model.
The key word in LUFS is loudness, not level. The measurement applies a K-weighting filter to the audio signal before calculating RMS energy. K-weighting is a frequency-emphasis curve that de-emphasizes very low frequencies (below about 100 Hz) and adds a slight shelf boost around 3–4 kHz — approximately where human hearing is most sensitive. The result is a measurement that correlates better with perceived loudness than a simple RMS or peak level measurement.
LUFS comes in three flavors relevant to podcast production:
- Integrated LUFS: The loudness measurement averaged over the entire program — the full episode length. This is the number that Apple Podcasts and Spotify normalize to. Target:
-16 LUFS. - Short-term LUFS: Measured over a 3-second sliding window. Useful for monitoring loudness during editing — you can see when a guest's voice is running consistently hotter or quieter than the host without waiting for an integrated pass.
- Loudness Range (LRA): The spread between the quietest and loudest passages in the program, expressed in LU (Loudness Units). For podcast dialogue, a well-edited episode typically sits in the 12–15 LU range. A very compressed, radio-style presentation might be 6–8 LU; a documentary with significant dynamic range between narration and ambient sound might be 18–22 LU.
The -16 LUFS target and why it's not negotiable
Apple Podcasts normalizes to -16 LUFS integrated. Spotify's normalization target for podcasts is also -16 LUFS, matching their music loudness normalization behavior. Both platforms apply this normalization at playback, not at ingest — they store whatever you upload, normalize at delivery.
The implication is counterintuitive to producers coming from music mastering, where louder masters have historically had a competitive advantage (the so-called loudness war). In podcast delivery, there is no competitive advantage to a hot master. If you deliver at -12 LUFS, the platform turns it down 4 dB. If you deliver at -20 LUFS, the platform turns it up 4 dB. In both cases, the listener hears approximately -16 LUFS. The loudness war is already over at the platform level.
What does matter is the shape of your dynamic range at delivery. A master at -16 LUFS integrated that has well-managed transients and a clean dialogue range will survive the platform's normalization pass with its character intact. A master that's been over-limited to squeeze up the integrated measurement will sound pumped and fatigued even after the normalization turn-down.
True-peak limiting: the -1 dBTP ceiling
Alongside integrated LUFS, true-peak level is the other measurement every podcast master needs to satisfy. The standard ceiling for podcast delivery is -1 dBTP (dB True Peak). Both Apple Podcasts and Spotify specify this, and it's consistent with EBU R128's guidance for streaming delivery.
True-peak metering is different from sample-peak metering. Digital audio samples represent amplitude at discrete points in time; the actual audio waveform between samples can exceed the measured sample value during digital-to-analog conversion. True-peak meters calculate inter-sample peaks — the maximum possible analog amplitude — typically by upsampling the signal by at least 4x before measuring. A file that measures -0.5 dBFS sample-peak might have true-peak values touching +0.5 dBTP or higher.
If your master crosses 0 dBTP, you get digital clipping in the DAC. This is audible as harsh distortion on transient peaks, and it survives the platform's normalization pass. A true-peak limiter set to a ceiling of -1 dBTP, placed last in your master bus chain, prevents this.
A practical master bus chain for podcast dialogue
Consider a two-host comedy podcast — two hosts in different cities, one a former radio host, one a content creator — recording weekly at around 12k monthly downloads. Their workflow involves multi-track stems exported from the recording session, edited for timing in a DAW, and run through a consistent master bus chain before export. A sensible chain for this kind of show:
- High-pass filter on each dialogue track: 80–100 Hz, 12 dB/octave. Removes room rumble and proximity-effect buildup from cardioid microphones.
- Dynamics per track: Light compression (4:1, slow attack ~20ms, medium release ~150ms) to control natural dialogue peaks without pumping.
- EQ per track: Presence boost around 2–4 kHz if needed; narrow cut around any resonant room frequency.
- Bus compressor: Very light glue compression on the stereo mix bus (1.5:1–2:1, slow attack, auto-release). Holds the mix together without audible pumping.
- Loudness metering: Check short-term LUFS during playback. Aim for consistent
-18to-16 LUFSshort-term during dialogue; don't ride the integrated toward-14. - True-peak limiter: Ceiling at
-1 dBTP, last in the chain. Not a brick-wall compressor — a limiter with a hard ceiling and fast look-ahead (typically 1–2ms).
This chain is deliberately not aggressive. Podcast dialogue doesn't need the density of a music master; it needs intelligibility, consistency, and headroom for the platform's normalization pass.
Where shows go wrong with loudness
We are not saying that normalizing to -16 LUFS is the only thing that matters in post-production. We are saying that ignoring it creates a predictable class of problems that are easy to avoid at export time and annoying to fix after distribution.
The most common issue: masters delivered at -19 to -23 LUFS. This happens when a host exports directly from their DAW without a loudness normalization pass, using default DAW settings that don't normalize at export. The platform applies upward normalization, turning the audio up 3–7 dB. If the original recording had clean headroom, this sounds fine. If it had noisy sections or inconsistent gain staging — common in a home-office recording environment — those sections are now noticeably louder at playback than the host expected when reviewing the edit.
The second common issue: over-limiting in an attempt to hit the target. A host measures their mix at -18 LUFS and applies a limiter at -1 dBTP with a 2–3 dB gain boost to reach -16 LUFS. The limiter's threshold is now triggering on every vocal peak. The result is a dense, fatiguing sound that the platform normalizes down exactly as much as the limiting boosted it — net effect: a squashed-sounding episode at the same target loudness as a clean-sounding one.
What loudness normalization doesn't fix
Loudness normalization is a delivery specification, not a production tool. It ensures your show sits at an appropriate volume relative to other content. It does not correct an uneven dialogue balance between hosts — if one host is consistently 6 dB louder than the other throughout the episode, the integrated target will be met but the listening experience will still be uneven. It does not fix a bad recording environment; a -16 LUFS master of a room with noticeable HVAC noise is a loud, well-normalized recording of HVAC noise.
The AES recommended practice for podcast delivery (AES TD1008) puts it precisely: loudness normalization targets address the delivery layer of the signal chain, not the production layer. Both layers have to be managed.
For shows that want to hand the normalized master to an editor or distribute directly, Rebel Audio's one-click export normalizes to -16 LUFS integrated with a -1 dBTP true-peak ceiling — the platform target baked into the export step so it doesn't get skipped.