What loudness should your film be mixed to?

Why every platform cares how loud your film is

Viewers hate reaching for the remote. Before loudness standards, television adverts were famously louder than the programs around them, and online video still jumps in level from one clip to the next when nothing controls it. Broadcasters, streaming services and audio apps now handle this in one of two ways.

The first way is a delivery specification. The broadcaster or streamer publishes a loudness number and a tolerance, and a file that falls outside it fails quality control. It is sent back to be fixed, or, under some contracts, corrected by the distributor with the cost charged back to the filmmaker. European broadcasters following EBU R128, US broadcasters following ATSC A/85, and Netflix all work this way.

The second way is playback normalization. The service accepts almost anything, measures each upload, and turns loud files down so they play at roughly the same level as everything else. Some services publish the level they normalize to. Others do not, and the number you see quoted is what engineers have measured from the outside.

This page and the tool keep those two kinds of number apart, because they carry different weight. Missing a delivery spec costs you a failed delivery. Missing a normalization level costs you some dynamics or some volume, and nobody rejects the file.

The units, in plain terms

Loudness meters report several numbers. Four of them matter for delivery.

• Integrated loudness is the average loudness of the whole program, from first frame to last, measured the way human hearing weighs different frequencies. It is the number every target on this page refers to.
• LUFS and LKFS are the units of integrated loudness. They are the same scale with two names: European documents say LUFS (loudness units relative to full scale) and US documents say LKFS (loudness, K-weighted, relative to full scale). A reading of -24 LUFS and a reading of -24 LKFS mean the same thing.
• LU is a loudness unit, used for differences. A change of 1 LU is the same size as a change of 1 dB, so a mix 3 LU too loud needs 3 dB of gain taken off.
• True peak, written dBTP, estimates the highest level the audio waveform reaches, including the peaks that fall between digital samples. An ordinary sample-peak meter can read lower than the true peak, which is why delivery specs ask for true peak.

All of the standards on this page measure loudness according to ITU-R BS.1770, the international recommendation that defines the measurement. Any meter that says it conforms to BS.1770 will give you numbers you can compare against these targets.

Why the meter ignores silence

BS.1770 uses gating, which means the meter leaves out very quiet passages when it calculates integrated loudness. An absolute gate drops anything below -70 LUFS, and a relative gate then drops passages more than 10 LU below the loudness of everything that passed the first gate. Without gating, a film with long silent stretches would measure far quieter than it sounds, and a mixer chasing the number would push the dialogue too loud.

Dialogue gating is different again

Netflix goes one step further and measures dialogue-gated loudness. The meter detects speech and calculates loudness over the passages where people are talking. The reasoning is that viewers set their volume by the dialogue, so the dialogue level is what should match across titles. A quiet drama and a loud action film can then have dialogue at the same level, with the action film's explosions sitting above it.

US broadcast television now works the same way for films. The July 2026 revision of ATSC A/85 says long-form content, which includes movies and episodic programs, is measured on its dialogue, while commercials and other short pieces are measured over the full program.

The practical consequence: a full-program reading and a dialogue-gated reading of the same film can differ by several dB. If you are delivering to Netflix or to US television, measure with a meter that supports dialogue gating before you change anything.

Published targets and observed levels

This table separates numbers that a named body publishes from numbers that are only observed. The tool uses the same values.

Every row can change. Broadcasters revise their delivery specs, streamers update their partner documents, and apps change their playback behavior without announcement. Treat the table as a starting point and confirm the current version with the destination before you deliver.

Destination by destination

EBU R128 broadcast

EBU R128 is the loudness recommendation of the European Broadcasting Union. It sets a target of -23 LUFS integrated, with a tolerance of plus or minus 0.5 LU for most programs, and a maximum true peak of -1 dBTP. R128 allows a wider tolerance of plus or minus 1 LU for live programs where hitting the target exactly is not practical. A finished film is not live, so aim for -23.0.

R128 is supported by a set of EBU technical documents that define the meter (EBU Tech 3341), the loudness range measurement (EBU Tech 3342) and practical guidance for production (EBU Tech 3343). Individual broadcasters add their own rules on top, such as limits on loudness range for certain channels. If a European broadcaster acquires your film, its delivery schedule is the document to follow.

ATSC A/85 broadcast in the United States

ATSC A/85 is a recommended practice from the Advanced Television Systems Committee. It sets a target of -24 LKFS for US television. In the United States, the CALM Act led the FCC to make A/85 mandatory for commercials, so that adverts play at the same average level as the programs around them. Networks apply the same target to programs through their own delivery specs.

A/85 itself gives a measurement tolerance of about plus or minus 2 dB around -24 LKFS, says the mix should not be aimed at either edge of that range, and asks for true peaks to stay below -2 dBTP. The tool uses those values. The July 2026 revision of A/85 also makes clear that a film or an episode is measured on its dialogue with a dialogue-gated meter, while commercials are measured over the full program. Networks publish their own delivery specs on top of A/85, and your network's spec has the final word.

Netflix

Netflix publishes sound mix specifications for its partners. The headline numbers are -27 LKFS, plus or minus 2 LU, measured dialogue-gated, with true peaks no higher than -2 dBTP. Netflix also publishes guidance on mix formats, stems and the deliverables that go with the mix, so read the whole document, not only the loudness line.

The -27 figure is a dialogue level. A film measured dialogue-gated at -27 will often read a little differently across the full program, and that is expected. It also sits 3 dB below the -24 LKFS dialogue-gated level that ATSC A/85 now gives for films on US television, so a Netflix mix and a US broadcast mix of the same film are two different files.

Apple Podcasts

If part of your adaptation is an audio drama, a podcast series or an audio trailer, Apple publishes audio requirements for podcasts. They recommend a loudness of about -16 LKFS, with a tolerance of plus or minus 1 dB, and true peaks no higher than -1 dBFS. Apple presents these as recommendations for creators rather than a rejection threshold, so the tool labels them as a published recommendation.

Spotify

Spotify publishes how its loudness normalization works for listeners. By default, playback is normalized to about -14 LUFS, and Spotify offers listeners louder and quieter settings as well. Spotify's guidance for masters is to keep true peaks below -1 dBTP, and below -2 dBTP if the master is louder than -14 LUFS, because louder masters lose more to lossy encoding.

Spotify says it turns louder tracks down and raises quieter ones, but how far a quiet track is raised depends on its peaks. Do not rely on the service to lift a quiet mix. Bring it up yourself, and check the true peak afterward.

YouTube

YouTube does not publish a loudness target for uploads. The figure of about -14 LUFS comes from engineers who measure what the player does: loud uploads are turned down to around that level, and quiet uploads are not turned up. The behavior has been reported consistently, but YouTube can change it at any time and has no obligation to tell anyone. You can check what happened to a particular video yourself: on a computer, right-click the player, open Stats for nerds, and read the Volume / Normalized line, which shows how far YouTube turned that video down.

The practical rule follows from that behavior. If your upload is louder than about -14 LUFS, YouTube turns it down and you gain nothing from the extra loudness. If it is quieter, it plays quieter than the videos around it.

TikTok and Instagram Reels

Neither TikTok nor Instagram publishes a loudness delivery spec or a documented normalization level. Many mixers use -14 LUFS integrated with true peaks below -1 dBTP as a working target for short vertical video, simply because it matches the other streaming services. Treat that as a convention. Many viewers scroll with the sound off, so captions and picture often carry the first seconds.

Cinema and festival screenings

Cinema does not use an integrated loudness target for features. A theatrical mix is made on a dubbing stage whose monitoring is calibrated to a reference level, and the theater plays the DCP (digital cinema package, the file set a cinema projector plays) with its own playback level set to the standard position. A commonly used calibration sets each screen channel so that pink noise at -20 dBFS plays at 85 dB SPL, measured C-weighted. The mix is judged by ear at that level.

This has two consequences for a book-to-film project. First, you cannot check a theatrical mix with this tool, so it returns an information result instead of a gain change. Second, a festival that asks for an online screener as well as a DCP is asking for a different file. Make a separate near-field mix for streaming, then measure that file against a streaming target.

A worked example: a Netflix mix that is 3 dB too loud

The tool opens with an illustrative Netflix delivery. The mixer measures the finished film with dialogue gating and reads -24.0 LKFS, with a true peak of -1.0 dBTP.

Step through it the same way the tool does.

1. Target. Netflix's published target is -27 LKFS, dialogue-gated, with a tolerance of plus or minus 2 LU.
2. Distance from target. -24.0 minus -27 is 3.0. The mix is 3.0 dB louder than the target.
3. Tolerance check. 3.0 is more than 2, so the mix is outside tolerance as it stands.
4. Gain change. Target minus measurement is -27 minus -24.0, which is -3.0 dB. Turn the whole mix down 3 dB.
5. True peak after the gain change. A gain change moves every peak by the same amount. -1.0 plus -3.0 is -4.0 dBTP.
6. Ceiling check. -4.0 is below Netflix's -2 dBTP ceiling, so no limiter is needed.

The tool reports a warning, not a failure, because one plain gain change fixes the file. The verdict reads "Turn it down 3.0 dB to hit -27 LKFS", the gain change shows -3.0 dB, the true peak after gain shows -4.0 dBTP, and the limiter row says No.

Notice what was not needed: no compression, no rebalancing, no remix. When a mix is only off in overall level, a single gain change across the whole program is the cleanest fix, because every relationship the mixer set between dialogue, music and effects stays exactly as it was.

A second example: a quiet YouTube upload

Now take an illustrative mix made for YouTube that measures -19.5 LUFS integrated with a true peak of -2.8 dBTP.

1. Target. About -14 LUFS, observed, not published by YouTube.
2. Distance. -19.5 is 5.5 dB quieter than -14.
3. What YouTube does. Based on observed behavior, nothing. YouTube turns loud uploads down but does not turn quiet ones up, so this video would play 5.5 dB quieter than most.
4. Gain change. -14 minus -19.5 is +5.5 dB.
5. True peak after the gain change. -2.8 plus 5.5 is +2.7 dBTP.
6. Ceiling check. +2.7 is 3.7 dB above the -1 dBTP working margin, and above full scale.

A gain change alone would clip the file. The fix is to raise the level through a true-peak limiter with its ceiling at -1 dBTP, so the loudest 3.7 dB of peaks are held down while the rest of the mix comes up. The tool reports "Raising it to -14 LUFS needs a limiter" and shows "Yes, 3.7 dB" in the limiter row.

That 3.7 dB is a useful warning sign. A few dB of peak limiting on short transients such as door slams or gunshots is usually inaudible. Several dB of limiting across dialogue or music can be heard as flattening or pumping. If the limiter is working hard, a better fix is to go back to the mix and bring down the few loudest moments by hand before raising the overall level.

A third case: inside tolerance already

An illustrative European broadcast master measures -23.2 LUFS with a true peak of -3.0 dBTP. It sits 0.2 dB under -23, well inside the plus or minus 0.5 LU tolerance, and its peaks are 2 dB below the -1 dBTP ceiling. The tool returns a pass. It still shows the +0.2 dB gain that would land exactly on -23, but no change is needed.

How gain and true peak move together

The arithmetic in the tool rests on one fact: a gain change moves loudness and peaks by the same number of dB. Turn a mix down 3 dB and both the integrated loudness and the true peak drop 3 dB. Turn it up 5.5 dB and both rise 5.5 dB.

That is why the tool always shows two numbers side by side:

• Gain change is target minus measurement.
• True peak after gain is your true peak plus that gain change.

If the second number is still under the ceiling, gain alone fixes the file. If it is over, the difference is how much peak reduction a limiter must do. Turning a mix down almost never needs a limiter. Turning a mix up often does.

What the tool does not check

The tool works from two numbers you type in, so it cannot hear the mix. Three things matter for delivery that no single loudness reading captures.

Loudness range. EBU Tech 3342 defines a measurement called loudness range, written LRA, which describes how far the loudness moves between the quieter and louder parts of a program. A film with whispered scenes and a loud climax can have a wide range. Some broadcasters set an upper limit on it for certain channels, and a very wide range on a phone speaker makes viewers ride the volume. If your delivery spec names a loudness range, read it from your meter and check it separately.

Dialogue clarity. A mix can hit its integrated target while the dialogue is buried under music. Integrated loudness is an average, and an average says nothing about whether a line of speech can be understood. Listen to the quietest dialogue scenes on small speakers as well as on the studio monitors.

Channel layout and file format. Delivery specs also name the channel layout (stereo, 5.1 and so on), the sample rate, the bit depth and the file type. A file can be perfect for loudness and still fail because it is the wrong format. Check those lines of the spec at the same time.

Making one film for several destinations

A book-to-film project rarely ends with one file. A typical set of deliverables might include a theatrical mix for festival screenings, a version for a streaming service, a version for online trailers and clips, and sometimes a broadcast version. Each one gets its own loudness target, and often its own mix.

Here is a practical way to organize the work.

1. Mix the main version first, to the strictest delivery you expect. For many independent films that is a streaming spec such as Netflix's, or a broadcast spec.
2. Keep the stems. Stems are separate audio files for dialogue, music and effects. A distributor may ask for them, and they let you make new versions without re-mixing from scratch.
3. Make an M&E track. An M&E (music and effects) track is the full mix without dialogue. Foreign distributors use it to dub the film into another language, and many sales agents ask for it.
4. Make derived versions from the main mix. An online version for YouTube is usually a near-field version of the main mix, raised toward about -14 LUFS through a limiter. Measure each version in this tool before you deliver it.
5. Label every file with its target. A file named for its destination and loudness avoids sending a -27 dialogue-gated mix to a channel expecting -23 integrated.

Common mistakes and how to avoid them

Measuring a section instead of the whole program. Integrated loudness is measured over the entire film. A loud action scene measured on its own says nothing about the film's integrated number. Always meter the full program.

Normalizing to peak instead of loudness. Many editing programs offer to normalize audio so its highest peak hits a certain level. That does not set loudness. Two files with the same peak can differ in loudness by 10 dB or more.

Reading sample peak instead of true peak. Sample-peak meters can read lower than true peak, and the difference grows after lossy encoding. Use a meter that reports dBTP.

Making the master as loud as possible. On services that normalize, a louder upload is turned down. All that remains of the extra loudness is the limiting used to create it.

Treating observed numbers as rules. The YouTube and social figures on this page are observed behavior and working conventions. They are useful targets, not promises from the platform. Check the platform's current help pages, and keep your master at a published standard so you can make new versions when the platforms change.

Confusing Netflix's target with a full-program target. Netflix's -27 is dialogue-gated. Mixing a film to -27 LUFS on a full-program meter will usually leave the dialogue at a different level from the one Netflix measures.

A delivery checklist

Run through this before any audio file leaves the mix room.

• Measure the full program with a BS.1770 meter: integrated loudness and true peak, and dialogue-gated loudness for Netflix or for a film going to US television.
• Enter the readings in the tool for each destination and apply the gain change it gives.
• Use a true-peak limiter only where the tool says the peaks would land over the ceiling.
• Measure again after every change. Do not assume the gain change landed where the arithmetic says.
• Listen to the loudest scenes and the quietest dialogue scenes after any limiting.
• Confirm the current published spec for each destination and keep a note of the version you mixed to.

Sources and rules this page relies on

• EBU R128, Loudness normalisation and permitted maximum level of audio signals, published by the European Broadcasting Union, with its supporting documents EBU Tech 3341 (loudness metering), EBU Tech 3342 (loudness range) and EBU Tech 3343 (production guidance).
• ITU-R BS.1770, Algorithms to measure audio programme loudness and true-peak audio level, published by the International Telecommunication Union, which defines the measurement every target here uses.
• ATSC A/85, Techniques for Establishing and Maintaining Audio Loudness for Digital Television, a recommended practice published by the Advanced Television Systems Committee, most recently revised in July 2026.
• The Commercial Advertisement Loudness Mitigation (CALM) Act and the FCC rules that implement it, for the US requirement on commercials.
• Netflix Sound Mix Specifications and Best Practices, published by Netflix in its Partner Help Center.
• Apple Podcasts audio requirements, published by Apple in its guidance for podcast creators.
• Spotify for Artists guidance on loudness normalization, published by Spotify.
• YouTube Help Center, published by YouTube, which does not set a loudness target for uploads; the YouTube figure here is observed behavior, labeled as such.

Specs and platform behavior change. Confirm the current version of each document with its publisher, and follow the delivery schedule your distributor sends whenever it differs from a general standard.