Podcast loudness: what LUFS to target

What loudness should a podcast target?

Target -16 LUFS integrated in stereo (the level Apple Podcasts recommends, within a 1 dB tolerance) and -19 LUFS in mono, with a true peak of -1 dBTP to avoid distortion when the file is encoded. Spotify normalizes to -14 LUFS, and YouTube to about -14 LUFS (an observed, de-facto level rather than an officially published figure). A single file at -16 LUFS and -1 dBTP behaves well across all of them.

What is LUFS, and how is it different from dB?

LUFS (Loudness Units Full Scale) measures perceived loudness over time, weighted for how the ear responds (K-weighting, defined by ITU-R BS.1770). Unlike a peak meter in dB, it reflects how loud something actually sounds. LKFS is the exact same measurement under the name used in North American broadcast.

• Integrated: the average loudness over the whole episode. This is the number platforms normalize to.
• Short-term: loudness over a rolling 3-second window.
• Momentary: loudness over a 400-millisecond window.

Per-platform targets (Apple, Spotify, YouTube)

Each platform publishes a reference, and a single well-leveled file covers them all. Master once to -16 LUFS rather than chasing every target separately.

• Apple Podcasts: -16 LUFS (LKFS), tolerance plus or minus 1 dB, true peak -1 dBTP.
• Spotify: -14 LUFS integrated.
• YouTube: about -14 LUFS integrated (an observed, de-facto playback level; YouTube does not officially publish a figure).
• Broadcast reference for context: -23 LUFS in Europe (EBU R128), -24 LKFS in the US (ATSC A/85).
• Commonly cited acceptable range: -14 to -18 LUFS.

Mono or stereo: -16 or -19 LUFS?

Export at -16 LUFS in stereo and -19 LUFS in mono. The 3 LU gap exists because a dual-mono file measures about 3 LU higher: the loudness meter sums the two identical, fully correlated channels before computing loudness, so the number is set by the meter's channel-summing convention, not by room acoustics. Delivering a mono file at -16 LUFS would make it noticeably too loud next to a stereo episode.

True peak: why -1 dBTP

True peak measures the real peaks reconstructed after digital-to-analog conversion, including inter-sample peaks a normal meter misses. Capping at -1 dBTP leaves the margin that prevents clipping during MP3 or AAC encoding. Apple asks for -1 dBTP or lower; Spotify recommends -1 dBTP, or -2 if the master exceeds -14 LUFS.

Loudness normalization vs peak normalization

Peak normalization sets the loudest sample to a value (often 0 or -1 dB) and does nothing for perceived volume. Loudness normalization adjusts gain to reach a LUFS target, which is the method platforms use and the most reliable way to keep volume consistent from one episode to the next. Setting a -0 dB peak does not fix loudness.

Prepare the voice before you set the loudness

Set loudness last. First clean the recording (noise and click removal in iZotope RX), correct the tone with EQ, then compress to even out the dynamics of the voice. Normalizing a noisy signal only amplifies its faults. The LUFS target applies to the finished mix, not the raw track.

Remote recordings add a wrinkle: guests arrive at different levels and often through lossy connections. Level each speaker separately before the loudness pass, and use spectral repair (RX) to recover high frequencies lost to call compression rather than relying on loudness alone.

Delivery checklist

Before you publish, confirm the finished episode against the targets.

• Integrated loudness around -16 LUFS (stereo) or -19 LUFS (mono).
• True peak at -1 dBTP or lower.
• Voice cleaned, EQ'd and compressed before the loudness pass.
• Consistent level between hosts, guests and any ad or music beds.
• Exported once and checked against a reference episode.

How to check your loudness on a meter

You read loudness on a loudness meter, not a peak meter. Load the finished mix, set the meter to the broadcast standard (ITU-R BS.1770), and let it run from the first word to the last. The integrated value it reports at the end is the number platforms normalize to. Momentary and short-term readings move constantly; the integrated figure is the one to match to your target.

DaVinci Resolve has a built-in loudness meter on the Fairlight page that reports integrated LUFS and true peak across the whole timeline, so you can measure without leaving the session. When a host DAW has no native LUFS readout, a dedicated meter plugin does the job; common neutral options include the free Youlean Loudness Meter and NUGEN's VisLM.

• Set the meter to ITU-R BS.1770 / EBU R128 mode before measuring.
• Let the integrated reading run across the entire episode, not just a loud section.
• Watch true peak on the same meter so you catch inter-sample peaks above -1 dBTP.

Setting the level: a concrete leveling pass

Once the voice is clean, EQ'd and compressed, set the final level in three steps. This is a reliable way to land on a target without guessing.

Re-measure after every gain change: a limiter or a gain move shifts the integrated reading, so the figure you trust is the one you read on the final export, not the one you aimed for.

• Put a true-peak limiter last in the chain with its ceiling at -1 dBTP so nothing exceeds the encoding margin.
• Raise or lower the overall gain until the integrated reading sits at your target (-16 LUFS stereo, -19 LUFS mono).
• Re-measure the whole episode end to end and confirm both the integrated value and the true peak before you export.

Common loudness mistakes

Most loudness problems come from measuring the wrong thing or in the wrong order.

• Reading a peak meter (dBFS) and assuming it tells you loudness. It does not; only an integrated LUFS reading does.
• Measuring only a momentary or short-term value instead of the integrated value over the full episode.
• Pushing for a louder number than the target, which the platform simply turns back down while keeping the squashed dynamics.
• Setting the loudness before cleaning, EQ and compression, so faults get amplified into the target.
• Forgetting the true-peak ceiling, so inter-sample peaks clip during MP3 or AAC encoding.