What LUFS measures
ITU-R BS.1770 defines algorithms to measure audio programme loudness and true-peak audio level; the current version is BS.1770-5 from November 2023. The loudness of a whole programme is measured with a level gate, which leaves the quietest stretches out of the average. The EBU notes that this gated measurement therefore captures the foreground loudness of a programme, and that it matches programmes with a wide loudness range better.
The result is expressed in LUFS, Loudness Units relative to Full Scale. BS.1770 and the ATSC practice call the same unit LKFS. Differences are in LU, and the ATSC document states that one unit equals a decibel: a file at −14 LUFS is 2 LU, or 2 dB, louder than one at −16 LUFS.
Meters show three time scales. Integrated loudness covers the whole file and is the number targets refer to. Short-term loudness uses a 3-second window and momentary loudness a shorter one; they show how the level moves. Loudness range describes how much the loudness varies over the programme.
Why loudness replaced peak levels
For decades audio was leveled by its peaks. The EBU's recommendation explains what went wrong: peak normalization led to considerable loudness differences between programs and channels, and those differences cause most viewer and listener complaints. A heavily compressed advert and a quiet drama can peak at the same level and sound very different. Measuring loudness and setting a common target fixes that, and leaves the peaks to a separate true-peak limit.
The targets
| Where | Target | Max true peak | Tolerance |
|---|---|---|---|
| Broadcast, Europe (EBU R 128) | −23 LUFS | −1 dBTP | ±1 LU where the target cannot be met, such as live |
| Television, U.S. (ATSC A/85) | −24 LUFS | −2 dBTP | about ±2 dB measurement tolerance |
| Podcast directory (Apple Podcasts help page) | −16 LUFS | −1 dBTP | ±1 dB |
| Music streaming default (Spotify help page) | −14 LUFS | −1 dBTP | normalized by the service |
EBU R 128 sets −23.0 LUFS for broadcast, with ±0.2 LU allowed for measurement error in quality control. ATSC A/85 recommends −24 LKFS for content delivered without metadata and asks that content not be aimed at either edge of its tolerance. The two platform figures come from their own help pages: the podcast directory asks creators for about −16 LKFS; the music service turns tracks to −14 LUFS by default and offers listeners a louder setting at −11 and a quieter one at −19. Its help page also asks for masters below −1 dB true peak, or below −2 dB if the master is louder than −14 LUFS. The EBU's streaming supplement confirms the general pattern: streaming services typically use a higher target than broadcast standards.
Albums, short clips and loudness range
Normalization does not have to flatten the difference between a quiet song and a loud one. The music service above says it normalizes an entire album at the same time, so the gain does not change between its tracks and a soft ballad stays softer than the single next to it. When the same tracks play shuffled or in a playlist, each one is adjusted on its own.
Loudness range describes the distribution of loudness within a programme. The EBU says it can be used to judge how much a programme's loudness varies and whether it will need further dynamic processing on its way to listeners. The EBU bases it on the 3-second short-term values and does not recommend it for programmes shorter than one minute, because there are too few data points. Short-form content such as adverts and promos has its own EBU supplement, R 128 s1. The EBU also allows a programme to be normalized below −23 LUFS on purpose, if that exception is clearly marked so nothing downstream turns it back up.
The gain arithmetic
Because LU are decibels, the gain a file needs is just the target minus the measured loudness. Negative means turn it down.
| Measured | To −23 | To −24 | To −16 | To −14 |
|---|---|---|---|---|
| −8 LUFS | −15 dB | −16 dB | −8 dB | −6 dB |
| −11 LUFS | −12 dB | −13 dB | −5 dB | −3 dB |
| −16 LUFS | −7 dB | −8 dB | 0 dB | +2 dB |
| −20 LUFS | −3 dB | −4 dB | +4 dB | +6 dB |
| −26 LUFS | +3 dB | +2 dB | +10 dB | +12 dB |
A normalizing platform applies this kind of gain for you. A master at −8 LUFS played on a service with a −14 target is turned down 6 dB. It ends up no louder than a gentler master, and turning it down does not undo the compression it was made with.
Turning up hits the peak limit
Raising a quiet file raises its peaks by the same amount. Take a podcast episode that measures −20 LUFS with true peaks at −3 dBTP. Reaching −16 LUFS needs +4 dB, which would push the peaks to +1 dBTP, over the digital maximum. The fix is a true-peak limiter or gentle compression before the gain, so the loud moments come down while the average comes up.
Decibels are logarithmic, so small numbers mean large changes in the signal:
| Change | Amplitude factor | Power factor |
|---|---|---|
| 1 dB | × 1.12 | × 1.26 |
| 3 dB | × 1.41 | × 2.00 |
| 6 dB | × 2.00 | × 3.98 |
| 10 dB | × 3.16 | × 10 |
| 20 dB | × 10.00 | × 100 |
The 1 dB margin under 0 dBTP that both the EBU and the podcast directory ask for is about 11 percent of amplitude. ATSC's −2 dBTP leaves about 21 percent, for the overshoot that lossy coding can add.
A practical workflow
- Edit first. Cut silence, false starts and long intros. The trim audio tool cuts the start and end with fades; the MP4 to MP3 tool pulls the sound out of a video recording.
- Measure. Use a meter that follows ITU-R BS.1770; the EBU requires one that also meets its Tech 3341 metering spec. Note the integrated loudness and the true peak.
- Pick the target. The strictest platform you publish to decides. A podcast heading for the directory above aims at −16; a broadcast delivery at −23 or −24.
- Limit, then gain. Bring peaks under control, apply the gain, and measure again.
- Export last. True peaks can change through encoding; the ATSC practice says to measure them before encoding and keeps extra headroom for that reason. The audio converter exports MP3, M4A and other formats at the bit rate you choose; the podcast directory's help page lists 128 to 256 kbps for stereo MP3.
None of the tools on this site measures LUFS. They handle the editing and format steps around the measurement, which happens in an audio editor or a dedicated loudness meter.
Frequently asked questions
What does LUFS stand for?
Loudness Units relative to Full Scale. It is the unit of loudness measured with the algorithm in ITU-R BS.1770. LKFS, the name used in BS.1770 itself and in the U.S. ATSC practice, is the same unit; the EBU uses LUFS. One LU of difference equals one decibel.
What LUFS should a podcast be?
Apple's podcast creator help page asks for about −16 LKFS with a tolerance of ±1 dB and a true peak no higher than −1 dBFS. Other directories may differ, so check each platform's own help page. If you also publish the audio elsewhere, mastering to the strictest target you need avoids two versions.
What is the loudness target for music streaming?
Each service sets its own. One large music service says on its artist help page that it adjusts tracks to −14 dB LUFS by default, with listener settings of −11 and −19. Masters louder than the target are turned down, so a very loud master gains nothing in level and keeps any distortion it picked up.
What is the difference between LUFS and dBFS?
dBFS describes the level of the signal itself, usually its peaks, relative to the digital maximum of 0 dBFS. LUFS describes perceived loudness averaged over time with frequency weighting and gating. Two files can peak at the same dBFS and differ by many LU in loudness.
What is true peak?
The highest level of the waveform in continuous time, which can be higher than the highest sample value. It is measured in dBTP. EBU R 128 sets a maximum of −1 dBTP in production; ATSC A/85 recommends keeping true peaks below −2 dBTP to leave room for later audio coding.
Sources
- ITU-R BS.1770: Algorithms to measure audio programme loudness and true-peak audio level (current version BS.1770-5, 11/2023)
- EBU R 128 (2023): target level −23.0 LUFS, ±1.0 LU where not practically achievable such as live, ±0.2 LU measurement tolerance, true peak not above −1 dBTP, LUFS equivalent to LKFS, gated programme loudness, loudness range from 3-second short-term values
- EBU R 128 s2, Loudness in streaming (streaming services typically use a higher target than broadcast standards)
- ATSC A/85:2026-07, Techniques for establishing and maintaining audio loudness for digital television (target −24 LKFS for content without metadata, about ±2 dB tolerance, true peak below −2 dBTP, one LKFS unit equals a decibel)
- Apple Podcasts for Creators: Audio requirements (about −16 dB LKFS ±1 dB, true peak not above −1 dB FS, recommended formats and bit rates)
- Spotify for Artists: Loudness normalization (tracks adjusted to −14 dB LUFS per ITU 1770; Loud −11, Quiet −19; masters below −1 dB true peak, below −2 dB if louder than −14 LUFS; albums normalized together)