“Can you make it louder? It sounds quiet next to our other videos.” That note usually comes after the client played the cut on YouTube, right after a music video. The mix was fine. It was made for a different destination.
Loudness stopped being a matter of taste a long time ago. Broadcasters and streaming services measure it as a number, and they either send the file back or change its volume to fit their target.
The numbers aren’t hard to learn. What trips teams up is that every destination uses a different one, and one mix can’t hit all of them.
In short
- Loudness is measured in LUFS over the whole program, and true peak in dBTP catches peaks a normal meter misses.
- European TV asks for -23 LUFS (EBU R 128), US broadcast for -24 LKFS (ATSC A/85) and Netflix for -27 LKFS measured on dialog.
- YouTube doesn’t publish a target, but measurements put its reference near -14 LUFS, and it turns loud videos down without turning quiet ones up.
- Make one master per destination instead of one mix that roughly fits everything.
- Measure the whole export before it goes out and write the numbers in the delivery email.
What LUFS and true peak measure
LUFS measures how loud a program sounds, averaged over its whole length. The method comes from ITU-R BS.1770, which weights frequencies roughly the way the ear does and leaves near-silent passages out of the average. LKFS, the name in US documents, is the same unit.
A peak meter answers a different question. It shows the highest sample. A heavily compressed mix can sit just under 0 dBFS all the way through and sound much louder than a dynamic mix with the same peaks.
A loudness meter shows two more numbers:
- True peak (dBTP) estimates the peak between samples, which is where the waveform ends up after conversion and lossy encoding. A file that never touches 0 on a sample meter can still clip when a platform re-encodes it.
- Loudness range (LU) describes how far apart the quiet and loud parts are. Specs use it as guidance more than as a pass or fail.
Broadcast and Netflix: the numbers that get files rejected
Broadcast specs are rules, and a file that misses them comes back. These are the ones a production company runs into most:
| Destination | Target | True peak |
|---|---|---|
| European TV (EBU R 128) | -23 LUFS | -1 dBTP |
| US TV (ATSC A/85) | -24 LKFS | -2 dBTP |
| Netflix | -27 LKFS on dialog | -2 dBTP |
EBU R 128 sets -23.0 LUFS for the whole mix. The current version (5.0, November 2023) allows ±1 LU only where hitting the target isn’t practical, as in live programs.
ATSC A/85 recommends -24 LKFS, and the CALM Act points US TV commercials to it. On long-form programs it measures the dialog rather than the whole mix, and it expects readings to vary by about 2 dB either way.
Netflix asks for -27 LKFS ±2 LU, gated on dialog, with peaks generally under -2 dBTP. That’s quieter than TV on purpose, so the dialog sits at a comfortable level and music and effects have room above it.
Whenever the video goes to a broadcaster or a platform with delivery rules, ask for the spec sheet. It beats any table on a blog, this one included.
YouTube, Instagram and the web: who turns you down
Most online platforms accept any level and change the playback volume instead. A loud mix gets turned down, and its squashed dynamics then play at the same level as everyone else’s cleaner mix.
These were checked in October 2026, and platforms change them without much notice:
- Spotify documents -14 LUFS as its default, measured with BS.1770, plus Loud (-11) and Quiet (-19) settings. It’s a music service, but it’s the best documented reference among the big ones.
- YouTube publishes no number. Mastering engineer Ian Shepherd measured its reference moving to -14 LUFS in September 2019, and noted that YouTube turns loud videos down but doesn’t turn quiet ones up.
- Instagram and TikTok publish no loudness target for uploads.
- The EBU’s streaming supplement (R 128 s2, 2023) lets a broadcaster raise its online version to between -20 and -16 LUFS.
So for a corporate video that lives on YouTube, LinkedIn and the client’s site, a mix around -14 to -16 LUFS with true peak under -1 dBTP is the common practice. Go much quieter and it sounds weak next to the video before it, since nobody turns it up for you.
One master per destination
A mix for TV and a mix for the web are two different files. Raising a -23 LUFS master by 9 dB to reach -14 means pushing it hard into a limiter, and the web mix played on TV comes out squashed and too loud.
What works on most jobs:
- Ask where the video will play before the mix starts. “Website and LinkedIn” and “a TV spot in two states” need different sessions, and the answer belongs on the deliverables list.
- Mix the strictest destination first. That’s usually the broadcast version, which has hard rules and wider dynamics.
- Make the web version as its own master pass, from the stems if you have them, instead of turning up the TV file.
- Name the files by destination, so nobody uploads the -23 master to YouTube by accident.
How to check the levels before you send
Measure the export, not the timeline. A plugin, a muted track or a render setting can change the result between the session and the file.
- Run the whole file through a BS.1770 meter, first frame to last. Integrated loudness only means something over the full length.
- Read true peak, not sample peak. In Premiere Pro, the export settings have a Loudness Normalization option with EBU R128, ATSC A/85 and a custom BS.1770 target, a true peak limiter and an optional loudness report.
- Listen once on a phone speaker and once on headphones. Dialog buried under the music on a phone is the complaint you’ll actually get, and no meter reports it.
- Put the measured numbers in the delivery email. “Web master, -14.2 LUFS, -1.3 dBTP” answers the question before the client’s agency asks. Our delivery email template has room for it.
The audio checks belong on the same pass as the rest of the final QC checklist, and our list of common editing mistakes covers clipping and level jumps inside the mix.
How this works in RecReview
In RecReview, AI Review runs on a version when you ask for it, or on every upload if automatic AI Review is on in the workspace settings.
It measures the integrated loudness and true peak of the whole file, and the finding shows the number it measured. It flags true peak above -0.5 dBTP, and loudness above -11 or below -22 LUFS, measured against web delivery at around -14 LUFS. A broadcast master at -23 shows up as quiet, which is expected for that file.
It also flags problems at specific moments: stretches of clipped audio, digital dropouts in the middle of speech, isolated clicks, and silences longer than five seconds away from the opening and closing seconds. Each one has a timecode and sits on the player bar next to the team’s and the client’s comments.
When the client writes “the music is too loud here” on one version, AI Review measures that passage in both versions on the next one and marks the note done when the level came down.
It doesn’t check channel layouts or dialog loudness, and it doesn’t certify a broadcast spec. Your loudness meter still signs off on the mix. AI Review is there to catch a web export that’s far too quiet, too hot or clipping before the client presses play.
