Meeting Broadcast Loudness Requirements

Overview

Loudness compliance is one of the few areas of production governed by actual regulation, and it is measured rather than judged, so a mix that sounds right can still fail. The good news is that the rules are simple once the units make sense, and meeting them does not require compressing the life out of a programme. This guide covers what the measurements mean, how to hit a target while preserving dynamics, and the specific mistakes that produce failures.

What You Need

• A loudness meter that measures to the standard your spec names
• The delivery specification, which states the target and tolerance
• A true peak limiter, used as a safety net rather than as a mixing tool
• The full programme, since loudness is measured across the whole thing
• Monitoring you trust at a consistent level
• A measurement report to accompany delivery, if required

Steps

Theatrical Film vs. Streaming Delivery & Automated QC

Understanding the distinction between room-calibrated theatrical mixing, web monitoring, and streaming platform QC prevents unwanted audio processing during distribution:

• Theatrical Room Calibration (85 dB SPL): Theatrical mixes are made to a calibrated room rather than to a loudness number. Netflix's sound mix spec sets no LKFS requirement for theatrical mixes and calibrates them to a reference of 85 dB SPL, and its near field mixes to 79 or 82 dB SPL, both from a -20 dBFS reference. ATSC A/85 uses 85 dB SPL, C-weighted and slow, only for rooms over 20,000 cubic feet, and steps down to 76 dB SPL for rooms under 1,500 cubic feet.
• Near Field Streaming (-27 LKFS) vs. Broadcast (-24 LKFS, -23 LUFS): Netflix's current spec asks for near field mixes at -27 LKFS ±2 LU, dialog-gated, measured to ITU-R BS.1770-1 across the whole programme. Where dialogue is under 15 percent of the programme, the whole mix is measured instead, at -24 LKFS ±2 LU. ATSC A/85, the recommended practice for US digital television, gives -24 LKFS. EBU R 128 sets -23.0 LUFS ±0.5 LU, or ±1.0 LU for live programmes. Other services publish their own figures, and the delivery spec you are given is the one that counts.
• Tolerance Is Not a Target: ATSC A/85 treats its ±2 dB as room for measurement variation and says content should not be aimed at the high or low side of it. EBU R 128 allows only ±0.5 LU. Aim at the number itself. What a platform does with a file outside tolerance, whether it rejects it or changes its gain, depends on the platform and is not set out in the documents read for this page.
• True Peak Safety (-2 dBTP to -5 dBTP): While nominal true peak specs specify -2 dBTP, many streaming automated processors crush peaks down to -5 dBTP. Setting your final limiter ceiling to -5 dBTP prevents unexpected distortion from automated platform limiters.
• Assembled Delivery Metering: Always measure the full, assembled final file (no reel breaks, 2-second heads/tails) using meters like iZotope Insight 2, Youlean Loudness Meter 2, or Waves WLM Plus before signing off on QC.

Pro Tips

• Mix first, normalise once. Chasing the meter while mixing produces flat programmes.
• Aim within ±1 LUFS of your target to pass automated streaming QC without aggressive gain-clipping.
• True peak limiting goes last in the chain and is a safety net, not a loudness tool. Set to -5 dBTP for safety against streaming platform gates.
• Measure the delivered file in the delivered channel configuration. Timeline measurements are estimates.
• Watch dialogue intelligibility separately from the loudness number. They are not the same thing.
• Keep the measurement report. It is evidence if a delivery is queried.

Tested: what integrated loudness counts, and what a gain change or a downmix does to it

We tested steps 1, 2, 5 and 6 on 2026-09-30 with FFmpeg 8.1.2's ebur128 filter, on synthetic band-limited noise programmes. This overlaps other guides' loudness tests and adds this page's own claims. The page's cinema and streaming target figures in the last section are not tested here. They were checked against the documents under Sources on 2026-10-05 and corrected on 2026-10-06.

Step 1 needs a limit. Adding a minute at about -50 LUFS to a minute at -23 LUFS left the integrated loudness at -23.2, unchanged, because the measurement gates out anything more than 10 LU below the running average. The page says quiet and loud passages both contribute. That holds for a passage a few LU down and not for very quiet ones, which is worth knowing when a programme has long near-silent stretches.

Step 2 holds. A gain of 6.02 dB moved integrated loudness by 6.10 LU, the small excess coming from the measurement's gate, and moved the loudness range from 7.1 to 7.0 LU, so a single offset keeps every balance decision, as the page says.

Steps 5 and 6 hold. The same 5.1 programme measured -16.1 LUFS as 5.1 and -17.8 after FFmpeg's stereo downmix, -1.7 LU apart, so a spec measured on one does not describe the other. With the front and surround pairs identical, the downmix measured -16.0 LUFS, and with the surround pair polarity inverted -31.3, a drop of 15.3 LU from phase cancellation alone. That is an extreme case, since the front and surround signals were identical, and real programmes fall somewhere between.

Validated for the measurements. Measured 2026-09-30 with FFmpeg 8.1.2 (ebur128) on synthetic noise, using its default downmix coefficients. Other downmix matrices and real programme material give other numbers. The delivery figures in the page's theatrical and streaming section come from the documents under Sources, not from this test.