Why dialogue is so quiet when you stream movies, and how to fix it

You turn the volume up to hear what two characters are whispering, and thirty seconds later an explosion rattles the windows. You turn it down. The next conversation is inaudible. Somewhere around the third round, you reach for the subtitles.

This is not your hearing and it is usually not a faulty TV. Quiet dialogue and loud action are what you get when a movie mixed for one kind of room is measured, delivered and played back in another. This guide explains the four mechanisms behind it, the loudness standards involved, and which fixes actually work.

The short answer

Movies keep a wide gap between dialogue and everything else. Streaming services such as Netflix deliberately preserve that gap. Then your TV squeezes a six-channel soundtrack into two small speakers, which takes away the one channel that kept the voices separate. Each step is reasonable on its own. Together they produce the remote-control routine.

1. The mix assumes a big room

Sound mixers work in studios calibrated to a reference level. Netflix’s published sound mix specification, which applies to its own productions, sets that reference at 85 dB SPL for a theatrical mix and at 79 or 82 dB SPL for a near-field mix, the version intended for home viewing. The spec describes a near-field room as similar in size to a typical living room.

The difference matters because a theatrical mix assumes a big, quiet, carefully tuned room where a whisper can sit far below an explosion and still be heard. The Netflix spec says theatrical mixes have no loudness requirement at all and may peak at the very top of the digital scale. A home mix has tighter rules, which is why it exists.

When a movie reaches you, the question is which of those two mixes it carries, and how far the studio adapted the home version from the cinema one.

2. The meter listens to the dialogue, not the explosions

Every modern loudness standard measures with the same algorithm, ITU-R BS.1770, which the International Telecommunication Union first published in 2006 and has revised several times since. It produces a single number for perceived loudness, written in LKFS or LUFS, which are the same unit under two names.

What differs between standards is what gets measured.

Netflix’s spec asks for an average of -27 LKFS, within 2 units either side, measured with dialogue gating. The meter only counts the moments when someone is speaking.

That choice is deliberate and it has a consequence. Because the explosions do not count towards the average, they do not force the mixer to turn the voices down to compensate. But nothing in the average stops the explosions being far louder than the voices either. The spec limits peaks to -2 dB true peak and recommends a loudness range of 4 to 18 units for a full mix, against 10 or fewer for the dialogue on its own. Eighteen units is a large gap in a living room.

The two kinds of number are not directly comparable, since one measures speech and the others measure everything. The practical reading is that broadcast standards control how loud a whole programme feels, while Netflix’s controls how loud the people feel and leaves room around them.

Netflix’s spec also has a fallback. When a 5.1 or Atmos mix contains less than 15 percent dialogue, it switches to a whole-programme target of -24 LKFS, because there is too little speech to measure.

3. Streaming services keep the gap on purpose

Broadcast television has long had a reason to keep things level. In the United States, the CALM Act, signed in December 2010 and enforced by the FCC from December 2012, requires commercials to be no louder on average than the programmes around them, using the ATSC A/85 method.

Streaming has no ad break to police in the same way, and Netflix’s spec makes its position explicit. Under best practices, it says Netflix is committed to safeguarding creative intent and refrains from compressing, limiting or altering audio mixes. It does normalise every delivery, raising or lowering the overall level so that one title plays back about as loud as the next.

Normalisation fixes the jump between titles. It does nothing to the range inside a title, because it moves the whole mix up or down by the same amount. The whisper and the explosion move together.

The spec does ask for dialogue to stay clear and intelligible, with an exception where a lack of clarity is part of the creative vision. That exception is honest, and it is also why some movies will always be hard to follow without subtitles.

4. Your TV folds six channels into two

A 5.1 surround mix has six channels: left, centre, right, two surrounds and a low-frequency effects channel. Dolby’s own speaker guidance says the centre speaker is primarily there to deliver dialogue, and that is how most mixers use it. In a surround system, the voices have a speaker to themselves.

Most people do not have one. TV speakers, laptops, tablets and many soundbars play two channels, so the device has to downmix. Netflix’s spec names ITU-R BS.775 as the international standard for this and spells out the method for its own stereo deliveries:

• Lower the centre channel by 3 dB and add it to both left and right.
• Lower each surround channel by at least 3 dB and add it to its own side.
• Optionally add the low-frequency channel at a much lower level, or leave it out.

Lowering the centre by 3 dB on each side is not a cut. Two copies at that level add back up to the original power, which is why the standard uses that figure. The problem is what joins it. The left, right and both surround channels, which in a cinema came from four separate places around you, now come out of the same two small speakers as the voices. Music and effects that used to surround the dialogue now sit on top of it.

The fold-down happens in different places depending on the setup. A service may send a dedicated stereo mix, the TV may downmix a surround stream itself, or a soundbar may do it.

What actually helps

These are settings rather than purchases, and the names differ by manufacturer.

• Reduce the dynamic range. Look for a setting called night mode, dynamic range compression, automatic volume or something close, on the TV, the soundbar or the receiver. It narrows the gap between quiet and loud. Broadcast formats have long carried the data to support this: the ATSC A/85 practice describes the dynamic range control built into Dolby’s AC-3 audio, with separate modes for different listening conditions.
• Tell the device what you actually have. If sound comes only from TV speakers, set the audio output to stereo rather than surround or auto, so the downmix happens once and deliberately.
• Try a speech or dialogue enhancement mode. Many TVs and soundbars have one, designed to make voices stand out from the rest of the mix.
• Use subtitles without guilt. Some movies mix dialogue low on purpose. Subtitles are the only fix that works on all of them.

The short version

• Dialogue is quiet because movies keep a wide gap between speech and action, and streaming services preserve it.
• Netflix measures loudness on dialogue alone, at -27 LKFS, and lets the rest of the mix peak well above it.
• Normalisation evens out the volume between titles, not within them.
• On two speakers, the downmix puts the music and effects in the same place as the voices.
• A dynamic range setting and the correct output mode fix most of it. Subtitles fix the rest.

The video side of the same problem, where a picture goes soft mid-scene because the stream dropped to a lower quality, has its own guide: why streaming picture quality drops in the middle of a movie. And if the part of the evening you would rather shorten is choosing the movie, that is the twenty-minute scroll.