How loud should a podcast be?

A podcast should be around -16 LUFS integrated loudness for a stereo file, with true peaks no higher than -1 dBTP; a mono file is usually aimed near -19 LUFS so that it plays at the same perceived loudness. Apple Podcasts normalizes to about -16 LUFS, while Spotify and YouTube normalize to about -14 LUFS.

The answer is a number, around -16 LUFS, and that number is useless without knowing what it measures and why anybody agreed on it. This is the reference page for both: what the unit is, what the platforms do with it, and what actually goes wrong at either extreme.

What LUFS measures

LUFS stands for Loudness Units relative to Full Scale. It is a measurement of how loud something sounds to a person, which is a different question from how large the signal is, and the difference is the reason the unit had to be invented.

A conventional peak meter reports the largest value the waveform reaches. That tells you whether you are about to clip, and almost nothing about loudness: a recording of a whisper with one door slam in it peaks exactly as high as a recording of shouting. LUFS instead applies a filter that weights frequencies the way the ear does, discounting the very low and very high ranges where we are less sensitive, then averages the result over time.

Three numbers come out of the standard measurement, and it is worth knowing which one people mean.

• Integrated LUFS is the loudness of the whole programme, start to finish, with the silent passages gated out so a long pause does not drag the figure down. This is the number a target refers to.
• Short-term and momentary LUFS are the same measurement over a three second and a four hundred millisecond window. They are what you watch while mixing, to see whether one segment is wandering away from the rest.
• True peak, in dBTP, is a separate limit. It estimates the highest point the waveform will actually reach after a digital file is reconstructed into sound, which can be above the highest sample stored in the file. It is a ceiling, not a target.

One more unit shows up in the same conversation: LU, a loudness unit, used for differences. Saying a file is 3 LU quiet means it sits three decibels of perceived loudness below where it should.

The targets the platforms use

The distribution platforms do not simply play what you upload. They measure it and adjust it toward a house loudness, so that moving between two shows does not mean reaching for the volume control. The published targets cluster tightly:

The spread between the podcast platforms is two decibels, which is small enough that one master serves all of them. -16 LUFS integrated with true peaks at -1 dBTP is the convention, and it exists because Apple published it first and the industry settled around it rather than because -16 is acoustically special. Broadcast sits ten decibels lower for a reason that does not apply to you: television has to accommodate drama with a genuinely wide dynamic range in a quiet living room.

Mono is the one place people go wrong on the arithmetic. A mono file played through two speakers is summed, and the standard measurement accounts for that, so a mono master aimed at about -19 LUFS ends up at the same perceived loudness as a stereo master at -16. Aim a mono file at -16 and it arrives noticeably loud.

Why there is a standard at all

Because the alternative was an arms race. When playback did no adjusting, any show that mastered itself louder than its neighbours simply sounded better on a shuffle, and everyone pushed until the only way to get louder was to compress the life out of the material. Music spent two decades on that, and the loudness normalization on modern platforms exists specifically to end it: if everything is played back at the same loudness, mastering hot buys you nothing and costs you dynamics.

For spoken word the practical argument is simpler. A podcast is listened to for forty minutes in a car, a kitchen, or on a train, on a phone speaker that may not have much headroom left. A predictable level means nobody has to ride the volume control between shows, or between your introduction and your interview.

What goes wrong above and below

Too loud. Uploading at -9 LUFS does not make you loud; the platform turns it back down to its target. What you keep is the damage done getting there, usually heavy compression or limiting that flattened the performance, now played at the same level as everyone else but with less life in it. If the true peaks were pushed to 0 dBFS you also risk audible distortion on playback, because lossy encoding can overshoot the peaks stored in the file, which is precisely what the -1 dBTP ceiling exists to prevent.

Too quiet. Some platforms turn a quiet file up and some do not, and neither outcome is good. When it is turned up, the noise floor comes with it. When it is not, your show is the one listeners have to reach for, and on a phone speaker at the upper limit it may not get loud enough at all.

Inconsistent between episodes is worse than either, and it is the common failure for shows without a levelling step. Episode four at -13 and episode five at -22 means a listener who set the volume last week gets it wrong this week, and the fix is not per-episode taste but a measurement applied to every file before it goes out.

Hitting the target is a measurement and one offset, not a craft. Normalize audio online covers doing it, and why your audio is too quiet covers the recordings where the number is not the real problem.

Common questions

CutNoodle is free and runs in your browser. Open the Simple Audio Editor and pick a file.