Loudness in Film Mixing: Understanding LUFS, Dialnorm and Delivery Standards

The mix is done, the client is happy, delivery is due. Then the QC report arrives: 'Programme loudness -22.1 LUFS, target -23 ±0.5 — fail.' Every film mixer has lived this. The problem is usually that a mix that sounds right by ear isn't speaking the language of the measurement standard. This article explains what LUFS/LKFS actually measure, what dialnorm does, and where the delivery standards (EBU R128, ATSC A/85, Netflix) differ, from a film mixer's point of view.

LUFS and LKFS: One Thing, Two Names

LUFS (Loudness Units relative to Full Scale) and LKFS (Loudness, K-weighted, relative to Full Scale) are two names for the same measurement; both come from the ITU-R BS.1770 algorithm. It first applies a K-weighting filter to the signal (approximating the ear's sensitivity), then averages the energy over time. LU is the relative unit: 1 LU equals a 1 dB difference. Unlike peak metering (dBFS), loudness describes how loud a programme sounds, not how high it reaches.

Three Measurement Windows

• ―Momentary (400 ms): for watching instantaneous swells. You see how a scene behaves during an explosion.
• ―Short-term (3 s): for reading the balance at sentence and scene level.
• ―Integrated: the average over the whole programme. This is the value delivery standards ask for.

Gating: Separating Silence and Dialogue

Integrated measurement uses a gate: very quiet passages are left out of the average. BS.1770-3 and later apply a relative gate (-10 LU), so a long quiet scene doesn't artificially drag the average down. Dialogue-gated measurement averages only the parts that contain speech, and for a film mixer this is the critical difference: cinema and streaming content carries a huge dynamic range between loud effects and quiet dialogue. That's why Netflix and ATSC A/85 anchor on dialogue, while EBU R128 measures the whole programme.

The Standards Side by Side

The values below summarise commonly published specifications. Always check the recipient's current spec before delivery; these values can change over time and by content type.

• ―EBU R128 (European broadcast): -23 LUFS integrated, ±0.5 LU tolerance; true peak no higher than -1 dBTP. Measures the whole programme.
• ―ATSC A/85 (US television): -24 LKFS, ±2 LU; dialogue-gated; true peak -2 dBTP.
• ―Netflix: -27 LKFS, ±2 LU, dialogue-gated; true peak -2 dBTP. Measured over the full programme with a dialogue-gated meter.

Netflix's lower target allows a wider dynamic range: dialogue sits at -27 while explosions can climb far above it. Delivering the same mix to EBU R128 usually means less dynamics, since the entire programme is pulled to -23. This is why a 'one mix for everywhere' approach rarely works.

What Is Dialnorm?

Dialnorm (dialogue normalization) is a metadata value embedded in a Dolby Digital (AC-3) bitstream, a number between -1 and -31. It tells the receiver what the programme's dialogue level is. The receiver (set-top box, TV, audio system) uses it to adjust output and level-match dialogue between programmes. Dialnorm doesn't change the mix itself; it only carries information for a correction applied at playback.

The most common error is setting dialnorm to a value you haven't measured. If the mix's dialogue is measured at -24 LKFS but dialnorm says -31, the receiver corrects the programme against the wrong level and the viewer hears inconsistent sound. The rule is simple: measure first, then write the value.

True Peak: Beyond the Digital Ceiling

Sample peak doesn't show what happens between samples; peaks reconstructed during D/A conversion or encoding can exceed 0 dBFS. A True Peak (dBTP) meter catches those inter-sample values through oversampling. Delivery limits typically sit between -1 and -2 dBTP, leaving margin for overshoots caused by codecs such as AAC or Dolby. A true-peak limiter at the end of the master bus is safe, but think of it as protection against explosions and music peaks, not as something that works on the dialogue.

A Practical Mixing Workflow

• ―Treat the dialogue as your reference throughout: set it to target first, then balance effects and music around it.
• ―Put a loudness meter on the mix bus (e.g., Dolby Media Meter, Nugen VisLM, iZotope Insight) and measure in dialogue-gated mode.
• ―When the mix is finished, measure the whole film start to finish in real time; judging a single scene is misleading.
• ―If you miss the target, revisit the dialogue bus and the effects balance instead of reaching for master gain.
• ―Enter dialnorm from the measured dialogue level.

Theatrical deliveries work differently from these broadcast and streaming standards; there a calibrated reference listening level is the basis, not a LUFS target. This article is about broadcast and streaming delivery.