The short answer
LUFS tells you how loud a track sounds, measured across its whole length. Streaming services use it to decide how much to turn your track up or down. True peak tells you the highest point the waveform reaches when it is played back. It decides whether the track distorts.
One number is about loudness. The other is about headroom. A track can be fine on one and wrong on the other, so a finished master needs both checked.
What LUFS measures
LUFS stands for Loudness Units relative to Full Scale. It is defined by ITU-R BS.1770, the standard that broadcasters and streaming services use. The meter first filters the audio to weigh it the way ears do: very low bass counts for less, and the upper middle counts for a little more. Then it measures the power of the audio in short blocks of 400 milliseconds.
For the whole track, quiet blocks are left out: first anything below −70 LUFS, then anything more than 10 LU below the level of what remains. What is left is averaged into one number: the integrated loudness. Shorter readings exist too. Short-term loudness covers the last 3 seconds, and momentary loudness the last 400 milliseconds.
One more thing worth knowing: LUFS and LKFS are the same unit with two names. The ITU standard writes LKFS, and the EBU and the AES both say the two are identical. One LU is one decibel. The full mechanics are in What LUFS actually measures.
What true peak measures
A digital file stores samples: tens of thousands of separate level readings each second. What reaches a speaker is the smooth wave those samples are rebuilt into. Between two samples, the rebuilt wave can rise higher than either of them.
So the highest sample is not always the highest point of the sound. A meter that reads samples (a sample-peak meter, in dBFS) can show −0.3 while the rebuilt wave goes over 0. Peaks like that are called inter-sample peaks. BS.1770 notes that overloads can happen even when no sample is at, or even close to, full scale.
A true-peak meter estimates the rebuilt wave. It oversamples, which means calculating extra points between the real samples (four times as many, in the standard’s method), and reports the highest one in dBTP. Unlike dBFS, a dBTP reading can be above 0.
Even a true-peak meter can read low
Four times oversampling is an estimate, not the exact wave. BS.1770-5 gives the worst case: a 4× meter can read up to 0.688 dB too low for content up to half the sample rate. That is one reason peak limits sit at −1 dBTP rather than at 0.
How the two numbers relate
The two numbers move independently. A heavily limited master can measure loud in LUFS with its true peak held just under its ceiling. A quiet, open acoustic recording can measure low in LUFS and still have sharp peaks near full scale, because a few loud transients barely change an average.
The gap between them has a name. PLR, the peak-to-loudness ratio, is the maximum true peak minus the integrated loudness. EBU R 128 s2 and AES TD1008 both define it. A track at −9 LUFS with a true peak of −1 dBTP has a PLR of 8 dB; a track at −16 LUFS peaking at −1 dBTP has a PLR of 15 dB. A smaller PLR means the loudest moments stand out less from the average: more of the track has been pushed up toward the peak.
PLR is not a score, and no standard sets a target for it. It is a way to see, in one number, how much room a master has left between its average level and its highest point. For how loudness changes over a whole song (quiet verse, loud chorus) the separate measure is loudness range (LRA), which the EBU defines and warns should not be confused with dynamic range or crest factor.
Which limits apply where
Each service publishes, or is measured to use, one loudness level, and some publish a peak limit. The loudness number is where playback lands. The peak number is the one to stay under.
- Spotify, -14 LUFS integrated, ceiling -1 dBTP (-2 dBTP for masters louder than -14 LUFS). Published by Spotify in its guide for artists, “Loudness normalization”.
- Apple Music, -16 LUFS integrated, ceiling -1 dBTP. Measured by third parties; Apple publishes −16 dB for Sound Check only in its podcast requirements, and asks for at least 1 dB of headroom in its Digital Masters guide.
- YouTube, -14 LUFS integrated, no published true-peak ceiling. Measured by third parties; YouTube publishes no loudness figure and no true-peak ceiling.
- Broadcast (EBU R 128), -23.0 LUFS integrated, ±0.2 LU, ceiling -1 dBTP. Published by the EBU in R 128 (2023). It is the level broadcast programmes are delivered at, not a player setting.
Some of those figures are measurements by third parties rather than numbers the service publishes, and the list says which. For distributors, AES TD1008 recommends that true peak not exceed −1 dBTP at the input of a lossy encoder. How each service handles a track that is too loud or too quiet is in streaming loudness targets.
Why streaming needs the extra headroom
Streaming services send listeners a lossy encode (AAC, Ogg Vorbis or Opus) not your WAV. A lossy encoder rebuilds the sound from a compressed description of it, and the rebuilt peaks can land higher than the peaks you delivered. BS.1770 lists bit-rate reduction among the things that can raise peaks.
How much depends on the encoder and its bit rate. AES TD1008 says a high-rate encoder (for example 256 kbps) may work with a limit as close as −0.5 dBTP, and that peaks rise more as the bit rate falls. A peak that clips inside the encode is in the stream every listener hears, and you never hear it before release.
Five things that are not true
- “No sample is over 0 dBFS, so it cannot clip.” The rebuilt wave can go higher than any sample. That is what true peak measures.
- “A louder LUFS reading makes my track louder on streaming.” With normalization on, a track louder than the service’s level is turned down by the difference. It does not end up louder than anyone else.
- “You must master to −14 LUFS.” Spotify suggests it, but nothing is rejected for missing it. The peak limit is the number that matters more.
- “LUFS and LKFS are different.” They are the same unit.
- “Loudness range is dynamic range.” The EBU, which defines loudness range, says it should not be confused with dynamic range or crest factor.
Checking both in practice
- Measure the exported file, not the mix in your session. Integrated loudness belongs to the whole finished track.
- Read true peak, not sample peak. If your DAW only shows sample peak, use a meter that oversamples. Our loudness checker shows integrated LUFS, true peak and loudness range together, and runs in your browser without uploading the file.
- If true peak is over the limit, lower the output ceiling of your final limiter, or turn the master down, and measure again. If the peaks are already distorted inside the file, turning it down will not undo that: the clipping detector shows where.
- Do not chase a LUFS number. Choose the loudness the music needs, keep true peak under the limit, and let each service set its own playback level.
Euphona’s mastering reports both numbers for every job: loudness, true peak and loudness range, measured before and after.
Questions people ask
Can true peak be above 0?
Yes. dBTP describes the rebuilt wave, not the stored samples, so a file whose samples all sit below 0 dBFS can read +0.5 dBTP or more. That file is likely to distort on playback or in a lossy encode.
Is −1 dBTP enough?
It is the limit Spotify, EBU R 128 and AES TD1008 use for most material. Spotify asks for −2 dBTP when the master is louder than −14 LUFS, and lower bit rates need more room. If a service you deliver to publishes its own limit, use that.
Which matters more, LUFS or true peak?
For playback level, LUFS, but the service sets the level for you, so it matters less than it seems. For distortion, true peak, and nobody fixes that for you after delivery.
Does true peak change the LUFS reading?
Barely. Integrated loudness averages the whole track, and a few short peaks hardly move it. That is exactly why the two are measured separately.
Keep reading