Loudness standards
Ten named presets ship in LevelAPI2, one per delivery practice and one per variant where a single document defines more than one. Pick the name printed on the spec your broadcaster sent you, and the target, the gating, the true-peak ceiling and what normalization anchors to are all configured in one call.
This page is the plain-language version. The API is in the documentation, and the measurements behind it are on the compliance page.
The question that comes up first, every time
LKFS and LUFS are two names for one measurement. LKFS, Loudness units relative to Full Scale, K-weighted, is the spelling used by the ITU and by ATSC: it appears in ITU-R BS.1770, in ATSC A/85, and in the national practices built on them such as ARIB TR-B32 and Free TV OP-59. LUFS, Loudness Units relative to Full Scale, is the spelling the EBU uses in R 128.
The algorithm is identical, the numbers are identical, and 1 LU equals 1 dB under either name. A programme measured at −24 LKFS is the same programme measured at −24 LUFS. Nothing needs converting. Each standard below is written the way its own document writes it.
Europe · broadcast, streaming, radio and cinema
R 128 is the European delivery practice, and the reason loudness normalization replaced peak normalization across European broadcast. It measures programme loudness over the whole item with a relative gate, and it fixes a true-peak ceiling so that normalization cannot produce clipping downstream.
The current practice, 2011 onwards
The default choice for European broadcast delivery. Programme loudness, gated, measured across the whole item.
Adapted streams
R 128 s2 keeps −23 LUFS for the broadcast chain but recognises that a stream adapted for an audience listening on phones and laptops sits higher. It gives a range of −20 to −16 LUFS for a broadcaster's own adapted stream; this preset takes the middle of it.
Wide dynamic range, anchor-based
For cinematic content, s4 describes normalizing on dialogue loudness rather than on the loudness of the whole mix, so that a film with a wide dynamic range is not pulled down by its loudest passages. This preset switches on the dialogue gate and points normalization at the dialogue meter.
Withdrawn · legacy material only
The first edition of R 128 gated at −8 LU. The EBU moved to −10 LU in 2011, to match the revision the ITU published, and the −8 LU version was withdrawn. It is here because archive material measured under it reads differently, and reproducing an old measurement is sometimes the job.
United States · and the one with legal teeth
In the United States, A/85 carries a legal driver. The CALM Act directs the FCC to hold commercial advertisements to the loudness of the programming around them, and 47 CFR 73.682(e) makes compliance with the A/85 recommended practice mandatory for commercials on television broadcast stations.
A/85 defines two recipes and they use opposite settings. Which one applies is decided by content length, not by preference, and picking the wrong one is the most common way to be out of compliance while believing you are inside it.
United States · programmes
Long-form programme material is anchored to dialogue. A/85 asks for a dialogue-gated measurement with the relative gate off, and says in as many words that a relative-level gate must not be used as a substitute for dialogue gating. This preset applies exactly that, and points normalization at the dialogue meter rather than the full mix.
United States · commercials and promos
Short-form material is measured on the full mix over its whole duration, with relative gating on and no dialogue anchor. This is the measurement that applies to the advertisements the CALM Act is about.
The long-form recipe has not changed across editions. A/85 was revised in 2011, 2013 and again in 2026, and the measurement it asks for on long-form content is the same in all of them. That is why there is one long-form preset rather than one per year: where the year does not change the measurement, the absence of a year is the message.
Dialogue gating is a measurement capability. The realtime adjuster does not perform it, so an A/85 workflow that depends on a dialogue anchor uses the measurement API.
Japan · broadcast
TR-B32 is the Japanese broadcast practice. It is gated the way R 128 is, with no dialogue anchor, and it differs from the European practice in its target and from the Australian practice in one decibel of true-peak headroom.
Japan · ARIB TR-B32 v1.5
ARIB publishes its own reference signals with expected values, and LevelAPI2 is measured against them. Those rows are on the compliance page, including the two signals that between them identify which relative gate a meter is actually applying.
Australia · broadcast
OP-59 is the Australian operational practice. It is configured like TR-B32 apart from the true-peak ceiling, which is 2 dB rather than 1. OP-48, which covers commercials certification, points at OP-59 for the measurement itself and adds no separate setting.
Australia · OP-59 Issue 4
OP-59's dialogue measurement is not automatic dialogue gating. The document describes an operator isolating dialogue by hand, and says that where dialogue cannot be reliably separated the full mix should be measured, with short-form material measured on the full mix over its total duration. So this preset leaves the dialogue gate off, which is what the document asks for.
The algorithm underneath all of the above
BS.1770 is not a delivery practice. It is the measurement algorithm that every practice on this page is built on, and it publishes no target level at all. It also writes the unit LKFS.
The revision matters, and it is the single largest source of two meters disagreeing. BS.1770-1 is ungated. BS.1770-2 introduced both the −70 LKFS absolute gate and the −10 LU relative gate. BS.1770-3 and -4 refined the true-peak and surround treatment without changing what a gated integrated reading returns, so a -2 and a -3 meter agree.
Both presets below carry a −23 target, which is the library's own default rather than a figure quoted from the ITU, so that the preset is usable on its own. Override it if you need a different one.
Ungated
Gated
One row per standard and per variant
| Standard | Region | Target | Relative gate | Dialogue gate | True peak | Preset constant |
|---|---|---|---|---|---|---|
| EBU R 128 | Europe | −23 LUFS | −10 LU | off | −1 dBTP | EBU_R128 |
| EBU R 128, 2010 edition | Europe | −23 LUFS | −8 LU | off | −1 dBTP | EBU_R128_2010 |
| EBU R 128 s2, streaming | Europe | −18 LUFS | −10 LU | off | −1 dBTP | EBU_R128_S2_STREAMING |
| EBU R 128 s4, cinematic | Europe | −23 LUFS | −10 LU | on | −1 dBTP | EBU_R128_S4_CINEMA |
| ATSC A/85, long form | United States | −24 LKFS | off | on | −2 dBTP | ATSC_A85_LONGFORM |
| ATSC A/85, short form | United States | −24 LKFS | −10 LU | off | −2 dBTP | ATSC_A85_SHORTFORM |
| ARIB TR-B32 | Japan | −24 LKFS | −10 LU | off | −1 dBTP | ARIB_TRB32 |
| Free TV Australia OP-59 | Australia | −24 LKFS | −10 LU | off | −2 dBTP | FREETV_OP59 |
| ITU-R BS.1770-1 | Algorithm | −23 LUFS | off | off | −1 dBTP | ITU_BS1770_1 |
| ITU-R BS.1770-4 | Algorithm | −23 LUFS | −10 LU | off | −1 dBTP | ITU_BS1770_4 |
Constants are shown without their LEVELAPI2_STANDARD_ prefix. The full names, the
numeric values and the calls that apply them are in the
API reference.
So a preset menu cannot go stale
A technician working to a delivery spec is not necessarily a loudness expert. They have a document with a name on it, and they need the meter to match. That is the whole design goal here: pick the name, and the guesswork is gone.
LevelAPI2 can list and describe its own presets at runtime. Three calls populate a dropdown directly from the loaded library, with the display name and a one-line description of exactly what each preset applies. Nothing is hardcoded in the host, so a library update adds standards to your product without a line of code changing, and a menu can never offer a standard the loaded build does not have.
A/85 itself warns that metering products label their presets differently, and that users should verify the algorithm rather than trust the label. A library that can state what a preset applies is the direct answer to that.
What was measured, and against what
All ten presets are verified to configure exactly what their own standard's document specifies. Each one is applied, read back, and compared parameter by parameter against the document it comes from. A mistyped constant fails that check.
Three are additionally verified against that body's own published reference audio: EBU R 128, ITU-R BS.1770 and ARIB TR-B32. Those are the standards bodies that publish conformance files with expected values and tolerances.
The other seven have no published reference set to measure against: only some standards bodies release conformance audio. The measurement path is the same either way. There is no per-standard measurement code anywhere, so the three verified standards exercise the engine all ten of them use.
None of this is a certification. Nothing has been assessed or approved by the EBU, the ITU, ATSC, ARIB or Free TV Australia. These are in-house measurements against reference files those bodies publish, with every expected value cited to its own clause so that any row can be checked independently. Dialogue-gated readings are reported rather than graded throughout, because no body publishes an expected dialogue-gated value for any file.