−24 LKFS: what it means, and why your export misses it

Every US public television distributor asks for the same number, and it is the one number an editing timeline cannot show you honestly. Here is what it measures, why your meters disagree with it, and what to check before the file leaves your machine.

Written for independent producers delivering to US public television · Checked September 2026

Have your mix measured — first audit free See how it is reported (PDF)

What the number actually measures

LKFS is loudness, measured to ITU-R BS.1770, averaged across the whole programme and weighted roughly the way hearing works. US broadcast practice, from ATSC A/85, sets the target at −24 LKFS, and distributors allow about ±2 around it. NETA asks for −24 LKFS ±2, measured over the entire programme.

Two words in that sentence do the damage: average and whole. It is not the loudest moment, and it is not a section you happened to check. A twenty-eight minute programme has one integrated loudness value, and that is the value your distributor reads.

Why your peak meter cannot tell you A peak meter in dBFS shows how close a sample came to the ceiling. Loudness is a different measurement entirely. A mix whose peaks sit at −6 dBFS can measure −28 LKFS — four decibels quiet, with the meters looking healthy all the way through. The two numbers do not move together, and neither one predicts the other.

Why “normalise to −24” in one click is not the end of it

Loudness normalisation moves the whole mix up or down until the average lands on target. That fixes one number and disturbs two others:

True peak. Raising a quiet mix raises every peak with it. Distributors that accept a loud moment still have a ceiling — NETA allows music and effects to touch −2 dBFS true peak for dramatic impact, not as a constant state.
Dialogue. A/85 anchors loudness on the dialogue, because that is what the audience follows. A programme carried by music can measure exactly −24 and still have dialogue buried under it. NETA writes this out as a shape: programme peaks between −12 and −8 dBFS, nominal −10, with the average around −20 dBFS.

Which is why a limiter added after you measured invalidates the measurement, and why a mix measured on the timeline is not the same as the file you actually export.

The three numbers to check before you send

Integrated loudness over the whole programme, from the exported file, not the timeline.
True peak, to see what the last stage of your chain did to the peaks.
Dialogue level, so you know the speech sits where a viewer expects it and not four decibels under the score.

Where it goes wrong, in practice Normalising clip by clip instead of the programme. Music beds mixed against a loud reference. A master limiter switched on after the measurement. Measuring one render and exporting another. Measuring only the first ten minutes because the export took an hour. Each of those produces a file that sounds right in the room and reads wrong at the station.

“It sounded fine on my speakers” is not a check

It is not carelessness — it is the wrong instrument. Your room, your monitor level and your ears are absolute about balance and useless about an averaged number defined by a standard. That number either comes from a meter that implements BS.1770, applied to the finished file, or it does not exist.

When I audit a master, loudness is measured across the whole programme, alongside true peak, clipping and the dialogue level, and each value is printed next to the value your distributor asks for. My measurement is validated against the EBU Tech 3341 and 3342 test set, so the number in the report is the number a station will get.

Want to know your number before the distributor does?

The first Delivery QC Audit is free — one free audit per production company, one file up to 60 minutes — with a written report in two business days, every item measured or checked on the picture.

Claim the free audit