EBU R128 Broadcast Loudness: A Plain Guide to the -23 LUFS Rule

Deliver a programme to a European broadcaster at the wrong loudness and it comes back. The rule is EBU R128: -23 LUFS integrated, true peak no higher than -1 dBTP. This guide explains the rule in plain words, how it differs from podcast and streaming targets, and how to hit it with a normal two-pass workflow.

The Broadcast — EBU R128 preset in the loudness normalizer applies exactly this target.

What R128 Changed

Before loudness rules, broadcasters normalized by peak, and adverts learned to sound louder without peaking higher — the loudness war your ears remember from old TV. R128 (built on ITU-R BS.1770, explained in the LUFS guide) measures perceived loudness of the whole programme instead. Compressors stopped being a cheat code: squash a programme and its integrated loudness betrays it.

The change deserves a moment, because it explains the rule's strictness. Under peak normalization, the only way to sound louder than the next advert was to compress harder — to shave every peak and fill the space with average level. Sound quality lost, viewers lost, and the broadcasters' own meters insisted everything was legal. R128 moved the goalposts from the loudest instant to the perceived whole, and the arms race ended within a few years: an advert compressed to within an inch of its life now measures exactly as loud as a gentle one at the same integrated level, and plays back at the same volume. The incentive to crush vanished.

R128 was published by the European Broadcasting Union in 2010 and adopted, country by country, across European television through the following decade. Its companions matter too: EBU Tech 3341 and 3342 describe the meters and the loudness-range measurement, and R128 supplements cover delivery details for files and short-form content. For a producer, though, the working rule fits on an index card — the next section's table.

The Core Numbers

The US equivalent, ATSC A/85, targets -24 LUFS with a -2 dBTP ceiling — close cousins, not twins. Delivering to both usually means two masters.

Read the tolerance line carefully, because it is tighter than podcasters are used to. A podcast at -15.2 against a -16 target is a shrug; a broadcast file at -21.8 against a ±1 window around -23 is a rejection. Broadcasters automate their ingest checks, and the machine does not shrug. Measure the delivered file, not the project, and aim for the centre of the window rather than its edge — meters differ by a tenth or two between implementations, and the centre forgives that.

Why -23? The Living-Room Logic

Podcasters meeting -23 for the first time often ask why broadcast sits so much lower than the -16 they know. Three reasons stack up. First, television is heard on real speakers in quiet rooms, where a lower average level still feels full. Second, broadcast programmes roam: a film under R128 may swing from a whisper to an explosion, and the average must leave headroom for the swing — a loudness range the -16 podcast world, built for earbuds in traffic, cannot afford. Third, history: -23 was chosen to sit near the average level of well-made legacy programming, so archives and new material could coexist without jarring jumps.

The practical consequence for you: a broadcast master is not a podcast master turned down. It is often a different mix decision — wider dynamics preserved, dialogue anchored to the lower target from the start. If you know a programme will live in both worlds, mix for the broadcast range and derive the podcast version with loudness processing, rather than crushing the podcast version down and hoping the dynamics come back. They do not.

How to Hit -23 LUFS

1. Mix for the target from the start. Dialogue anchored near the target makes the final step trivial.
2. Measure the whole programme. Integrated loudness is a full-file number; a section check is not a delivery check.
3. Normalize two-pass. Measure, apply the exact correction, hold true peak at -1 dBTP, re-measure. The workflow mirrors the podcast guide with a different target.
4. Ship the report with the file. Broadcasters increasingly expect the measured values; a normalizer that emits a per-file report (as LoudNorm does) saves a round trip.

A Worked Example: Delivering a Documentary

A 52-minute documentary is mixed in a small facility and exported for a European broadcaster. The first full-file measurement reads -25.8 LUFS integrated, true peak -3.4 dBTP, LRA 14 LU.

The LRA is fine for the genre — a documentary may roam. The loudness is 2.8 LU under target, outside even the generous edge of the window. The fix is one clean gain of +2.8 dB, except that the gain would carry the true peak to -0.6 dBTP — over the ceiling. So the delivery pass applies +2.8 dB with a true-peak limiter watching the handful of peaks (a door slam, two music hits) that cross. The re-measure reads -23.1 LUFS, -1.0 dBTP: inside the window, ceiling respected, dynamics untouched below the peaks. The report — integrated, true peak, LRA — goes in the delivery folder with the file, and the broadcaster's ingest accepts it on the first pass.

Notice what did not happen: nobody re-mixed, nobody compressed the programme flatter, nobody chased -23.000. The mix was already shaped for broadcast; the delivery pass is measurement, gain, ceiling, proof.

A Delivery Checklist You Can Steal

Before any file leaves for a broadcaster, run this list in order:

1. Confirm the spec in writing. Which broadcaster, which version of their delivery document, file format and naming rules. Specs differ in details that matter — sample rate, channel layout, metadata.
2. Export the programme master. Nothing after this export except the delivery pass itself.
3. Measure the export. Integrated loudness, true peak, LRA — all three, off the delivered file.
4. Correct in one pass. Gain to the centre of the loudness window, ceiling at -1 dBTP, re-measure.
5. Check the short-form caps if the content is an advert, promo or trailer.
6. Listen to the loudest minute and the quietest minute on ordinary speakers. Meters certify numbers; ears catch the mislabelled file, the wrong episode, the render with the missing channel.
7. File the report with the master. If a query arrives in three weeks, the report answers it in three minutes.

Most rejections die at step 1 or step 3: a spec nobody re-read, a file nobody measured. The loudness itself is rarely the hard part — the hard part is the discipline of measuring the exact artifact you ship.

R128 and A/85 Side by Side

Delivering one programme to both continents? Make the A/85 master first — its peak ceiling is stricter — then derive the R128 version with a +1 LU adjustment and a fresh measurement. Going the other way invites a peak failure in the US check.

Short-Form Content: Adverts and Promos

R128's programme rule alone would let a 20-second advert be ferociously loud in its loudest moments while averaging -23. So short-form content carries extra constraints on maximum momentary and short-term loudness — in plain terms, brief content may not spike far above the programme level around it. If you produce promos, trailers or adverts for European broadcast, the headline -23 is necessary but not sufficient; check the delivery spec of the specific broadcaster, which states the short-form caps it enforces. Podcasts promos in a feed face no such rule — this is purely a broadcast delivery matter, and it catches podcasters moonlighting in TV by surprise.

Podcasts vs Broadcast: Do Not Mix the Targets

A podcast master at -16 LUFS is 7 LU too hot for R128 delivery, and a broadcast master at -23 sounds quiet in a podcast queue. The formats share the measurement, never the target. If your programme lives in both worlds, keep two masters and label them clearly.

Label them like you mean it. File names such as episode12_broadcast_R128_-23.wav and episode12_podcast_-16.mp3 have saved more producers than any meter. The failure mode this prevents is real and common: the wrong master attached to a delivery form at deadline o'clock, discovered days later in an automated rejection email. Two masters, two names, two folders — and a delivery checklist that reads the loudness from the file, not from the filename, before anything is sent.

Common Mistakes

Delivering the podcast master. Seven LU hot is not a rounding error; automated ingest will flag it instantly. The masters are different files, made deliberately.

Measuring a segment. "The interview section sits at -23" is not a measurement of the programme. Integrated loudness belongs to the whole file, including the loud titles and the quiet credits — which is why it is measured last, on the export.

Ignoring the true-peak ceiling until the end. If the mix was built hot, the delivery pass may have to choose between the loudness window and the peak ceiling. Build in headroom from the first mix session; ceilings are cheap early and expensive late.

Assuming -23 travels. An R128 master sent to a US broadcaster is 1 LU over the A/85 target and 1 dB over its stricter peak ceiling — probably rejected, definitely re-measured. And a file made for European broadcast dropped into a podcast feed will sound politely quiet. Every destination gets its own delivery check.

Treating LRA as a pass/fail. R128 expects LRA to be reported and appropriate to the genre; there is no single legal range. Do not flatten a film to satisfy a number nobody set.

Frequently Asked Questions

Is R128 a law? It is a recommendation by the European Broadcasting Union that broadcasters adopt into their delivery contracts — which makes it binding in practice for anyone delivering to them. Several countries also reference loudness rules in regulation. For a producer, the distinction rarely matters: the ingest check enforces it either way.

What measurement standard does R128 use? ITU-R BS.1770, with gating — the same family of measurement behind every LUFS figure in podcasting. A BS.1770 meter that reports gated integrated loudness, true peak and LRA gives you the full R128 picture.

My programme is exactly -23 but was rejected for peaks. How? Integrated loudness and true peak are separate requirements; a file can pass one and fail the other. Peaks over -1 dBTP — often just a few moments — fail delivery on their own. The true peak guide covers the fix: limit the peaks, re-measure both numbers.

Can I normalize a finished programme to -23 with an online tool? Yes, for the loudness and peak requirements: the two-pass method with a -1 dBTP ceiling is the same one used in facilities. LoudNorm's Broadcast — EBU R128 preset applies exactly -23 LUFS and -1 dBTP and returns the per-file report to ship with the delivery. What a tool cannot do is judge genre questions — whether your LRA suits the programme type is an editorial call.

What about loudness for European radio and podcasts? Radio in much of Europe uses its own targets (often also in the -23 family for EBU radio, but station specs vary), while podcasts follow the -16/-19 conventions covered elsewhere in these guides. Always deliver to the written spec of the destination in front of you.

References

• https://tech.ebu.ch/publications/r128 — EBU R128, the recommendation itself, with its supplements on loudness range and delivery.
• https://www.itu.int/rec/R-REC-BS.1770 — ITU-R BS.1770, the measurement algorithm R128 adopts.
• https://artists.spotify.com/ — Spotify for Artists, for contrast with the -14 LUFS streaming target.