For years, moving from one channel to another, or from a programme to its ad break, meant a hand on the volume. Levels jumped because every source arrived with its own idea of loudness. Modern broadcast fixed this with an approach that surprises you when you first meet it: instead of flattening everything at encode, it carries a metadata value that lets the decoder do the levelling. That value is dialnorm.
Dialnorm, a metadata value and not a processor
Dialnorm, short for dialogue normalization, is a number written into the coded audio stream, typically in Dolby metadata. It states the average dialogue loudness of the programme, expressed in LKFS, the loudness unit of broadcast. Its value runs from minus 1 to minus 31 dB in AC-3.
The key point bears repeating because it is always forgotten: dialnorm does not touch the signal at encode. It compresses nothing, limits nothing. It is the decoder, at the viewer’s end, that reads this value and applies attenuation to bring every programme to a fixed playback target. Two programmes with correct dialnorm therefore come out at the same perceived level, whatever their original volumes. The target is standardised: minus 24 LKFS in the United States under ATSC A/85, minus 23 LUFS in Europe under EBU R128.
Why dialogue is the reference
Choosing dialogue as the anchor is no accident. On television, speech carries the meaning, and it is on speech that the ear sets its sense of loud or quiet. An action film and a news bulletin have very different dynamics, but if their dialogue comes out at the same level, the viewer leaves the volume alone. The measurement relies on speech detection, sometimes called dialogue intelligence, which isolates the spoken passages to estimate their loudness without being fooled by music or effects.
AC-3, AC-4 and the broadcast chain
Dialnorm lives in the metadata of broadcast codecs. You find it in legacy AC-3 and its big brother Dolby E used in production, and it stays present in next-generation systems such as AC-4 and MPEG-H, where loudness management gets even finer. The value is set at mastering or encode, once the programme’s integrated loudness has been measured.
Two schools coexist. The North American model often leaves each programme at its own loudness and sets dialnorm to describe it faithfully: the decoder handles the levelling. The European model prefers to normalise every programme to minus 23 LUFS in production, then set dialnorm accordingly. Either way, the metadata must reflect the reality of the signal, or the whole structure collapses.
The traps of dialnorm
The number one danger is the wrong value. If dialnorm claims minus 24 while the real dialogue comes out at minus 18, the decoder applies the wrong correction and the programme barges in too loud. That is exactly the level-jump mechanism that laws ended up regulating, the CALM Act in the United States in 2010 and the spread of EBU R128 across Europe.
A classic trap comes from inherited settings. For a long time, some channels left dialnorm stuck at minus 31, the most permissive default, which amounted to switching off normalisation and pushing the sound to the maximum. Keep in mind too that dialnorm drives the dynamic range control profiles, the Line and RF modes, which decide how much a small speaker gets squashed. A badly set value therefore knocks more out of place than just the average level.
In practice
- Measure the integrated loudness of each programme with a compliant loudness meter, not a plain peak meter.
- Set dialnorm to match that reading, or normalise to minus 23 LUFS first and set dialnorm to the target.
- Check that the metadata survives the whole chain, from production to transmitter, without being wiped by an intermediate device.
- Never let a default value stand in for a measurement.
The logic matches streaming and platforms, which apply the same playback-side normalisation idea, as with podcast loudness. Everywhere, the goal is the same: a steady level from one source to the next, with no action from the listener.
My take
Dialnorm is an elegant idea with a single flaw, it depends entirely on the honesty of whoever sets it. Measured well, it makes channel-hopping silent and the listening experience seamless. Set badly, it becomes the weapon behind the level jumps everyone hates. The good news is that measurement tools are now reliable and built in everywhere, and a correct value costs only a little rigour. It is one of those invisible settings you only notice the moment they go missing.
