True Peak Limiting and LUFS: Loudness for Streaming and Clubs

The limiter is usually the last plugin on the master, and the one people blame (or thank) for how loud a track ends up. Used carefully it's almost invisible. Pushed too hard it's the reason a song sounds flat and tiring, and somehow smaller even though it's louder. To use it well you need two things: an understanding of how a brickwall limiter works, and a clear idea of how loud your music actually needs to be.

Limiter vs compressor

A brickwall limiter is a compressor taken to the extreme: a ratio so high it's described as ∞:1, a very fast attack and an output ceiling the signal isn't allowed to pass. Modern limiters also use lookahead: they delay the audio by a few milliseconds so they can "see" peaks coming and reduce them smoothly instead of chopping them off.

A compressor shapes dynamics. A limiter mainly protects a maximum level, and in doing so lets you raise the overall loudness.

• Input gain or threshold: how hard you drive the limiter. More drive means more loudness and more gain reduction.
• Ceiling (output): the maximum level the output can reach.
• Release: how fast the limiter recovers after reducing a peak.
• Lookahead: how far ahead it reads the signal, usually 1–5 ms.
• True peak / oversampling: detection of peaks between samples.
• Style or algorithm: from transparent to aggressive and punchy.

Inter-sample peaks: why sample peaks lie

Digital audio is a series of samples. A normal peak meter shows only the level of those samples. But when the audio is converted back to a continuous waveform, by your interface's converter or by a playback device, the smooth curve between two samples can rise above both of them. These are inter-sample peaks. A master whose highest sample is −0.1 dBFS can reconstruct to +1 dB or more on dense, bright material, and clip the converter.

The same thing happens during lossy encoding. Streaming services convert your WAV into formats like AAC or Ogg Vorbis, and the encoding slightly reshapes the waveform. Peaks that were safely under 0 dBFS can come out above it.

A true peak limiter oversamples the signal (typically 4× or more) to estimate the reconstructed waveform and keeps that under the ceiling. The reading is shown in dBTP, decibels true peak. Turn true peak mode on for every final master.

Choosing a ceiling: −1.0 dBTP vs −0.3 dBTP

Spotify, for example, recommends keeping masters below −1 dBTP, and below −2 dBTP if the master is louder than −14 LUFS integrated. If you only make one version, −1.0 dBTP is the sensible choice.

Release: the setting that changes the most

A short release recovers quickly, so the track gets louder, but it can distort, especially bass-heavy music where the gain starts following low-frequency waveforms. A long release is cleaner but can cause audible ducking: the whole mix dips every time a loud kick or snare hits. Automatic release is a very good starting point on most limiters. If you set it manually, listen to the bass and to sustained sounds like pads and vocals; if they wobble in level, the release doesn't suit the material.

How much limiting is too much?

Watch the gain reduction meter, but trust your ears more. For dense modern mixes, roughly:

• 1–3 dB of reduction on peaks is usually very transparent.
• 3–6 dB starts to be audible on many tracks, though some genres take it well.
• Beyond that you're trading clarity for loudness, and you should know exactly why.

Listen for the snare and kick losing snap, grit in the bass, the vocal being pushed back as the beat hits, and reverb tails becoming strangely loud. If the limiter reacts constantly to one element, like a kick 6 dB hotter than everything else, fix it in the mix rather than leaning on the master.

What LUFS measures

Peak meters tell you the highest point of the waveform, not how loud music feels. Two songs can both peak at −1 dBTP and differ in loudness by 8 dB. LUFS, Loudness Units relative to Full Scale, is defined in ITU-R BS.1770 and used by the EBU R128 broadcast standard. It applies a frequency weighting (K-weighting) that roughly matches human hearing and gates out silence when measuring a whole track. One LU equals one dB, so −10 LUFS is 4 dB louder than −14 LUFS.

• Integrated LUFS: the average loudness of the whole track. This is the number streaming services use.
• Short-term LUFS: a sliding 3-second window. Useful for comparing chorus and verse.
• Momentary LUFS: a 400 ms window that reacts to quick changes.
• Loudness range (LRA): how much loudness varies through the song.
• True peak (dBTP): the highest reconstructed peak, as explained above.

Loudness normalization on streaming platforms

Most streaming platforms measure every track and adjust playback so songs play at similar loudness. If your master is louder than the reference level, it gets turned down. Some platforms also turn quieter tracks up, within limits. The figures below are the commonly cited reference levels; platforms change their systems from time to time, so check current documentation before treating any number as final.

−14 LUFS vs −8 LUFS: which should you master to?

−14 LUFS isn't a rule, it's a playback reference. Many commercial pop, hip-hop and electronic masters sit between −9 and −6 LUFS integrated. On streaming they're simply turned down, and they keep their dense, compressed character.

What normalization changes is the reward for extreme loudness. Say one master is at −8 LUFS and another at −14 LUFS. Spotify turns the first one down by 6 dB, so both now play at the same level. If the −8 version was squashed to get there, it may now sound smaller and flatter than the dynamic one, with no loudness advantage left. The −8 master still makes sense for club play, where nothing is normalized and a DJ wants the track to hit as hard as the record before it.

My approach: master for the sound the genre needs, not for a number. Then check that the result still holds up when turned down to −14 LUFS. If it sounds worse than a quieter version at matched loudness, back off the limiter.

Compare at matched loudness, always

This is the most important habit with a limiter. A limited master is louder than the unlimited mix, and louder always sounds better at first. Use your limiter's unity-gain audition feature if it has one, or lower the output while comparing so both versions sit at the same integrated LUFS. You'll often find the less limited version actually sounds better. Do the same with reference tracks: normalize everything to the same LUFS before judging tone, punch and width.

Common loudness myths

• "Spotify rejects tracks louder than −14 LUFS." It doesn't. They're just played back at a lower level.
• "Louder masters get penalized." Not beyond being turned down. The real risk is that extra limiting and encoding distortion make them sound worse once levels are equal.
• "Quiet masters will sound weak." Within reason, normalizing platforms bring them close to other tracks. A very quiet master may not be turned up fully, though.
• "LUFS tells you if a mix is good." It only measures loudness. A great mix and a bad one can read exactly the same.

A simple limiter workflow

1. Put the limiter last on the master chain, before dither.
2. Set the ceiling to −1.0 dBTP and enable true peak detection.
3. Start with automatic release.
4. Raise the input until the loudest sections show 1–3 dB of gain reduction.
5. Measure integrated and short-term LUFS against your references.
6. Compare with the limiter bypassed, at matched loudness.
7. If you need more level and it starts to suffer, go back to the mix or add gentle clipping instead of pushing harder.