Loudness Targets for Video and Why Platforms Turn It Down
A 42 minute interview film I mixed last spring came back from the client with one note: the music is louder than the people. In my room it was not.

A 42 minute interview film I mixed last spring came back from the client with one note: the music is louder than the people. In my room it was not. On her phone, held at arm's length in a kitchen, it absolutely was, and the uploaded version on YouTube sat noticeably quieter than the trailer that played before it.
Both problems came from the same place. Every major platform measures the loudness of your upload and applies its own gain change before anyone hears a frame, and almost nobody mixing video has looked at what number their file actually arrives at. Here are the targets that matter, how to read them inside an editor rather than a mastering suite, and why the popular advice to hit minus 14 LUFS makes mixes worse more often than it helps.
What normalization actually does to your upload
LUFS stands for loudness units relative to full scale. Unlike a peak meter, it follows a frequency weighting curve built to match how we hear, and the integrated reading averages the whole programme, ignoring silence. That single number is what the platform reads.
What happens next is simpler than people assume. The platform does not re-mix anything. It applies one gain offset to the entire file, the same number of decibels from the first frame to the last. YouTube turns loud uploads down and leaves quiet ones alone. Apple Podcasts and Spotify, with normalization switched on, will also raise quiet files toward their target.
You can watch it happen. Right click a YouTube video, choose Stats for nerds, and look at the content loudness line. It shows the offset applied in decibels. On a file I delivered at minus 11 LUFS it read minus 3.1 dB, which is the platform quietly undoing three decibels of limiting I had spent an hour applying.
The part that trips up editors coming from music: because the offset is a single gain change, arriving pre-squashed buys you nothing. A heavily limited mix and an open one both get pulled to roughly the same level. The limited one just has less left between a whisper and a shout once it gets there.
Meter the dialogue, not the whole timeline
Resolve, Premiere and Audition all ship a loudness meter, and most editors never open it. Put one on the master bus and leave it there. You want three readings from it: integrated for the whole programme, short term for the last three seconds, and true peak.
Then stop thinking about the master and start thinking about speech. The anchor of a video mix is the dialogue, because that is what a viewer sets their volume knob by. I aim for spoken word sitting around minus 18 to minus 16 LUFS short term during normal delivery, and I keep music beds roughly 8 to 12 LU below that while anyone is talking. Get those two relationships right and the integrated number lands close to where it needs to be without any help.
The order I work in on an interview edit: clip gain each speaker so their waveforms look alike before touching a plugin, one compressor at about 3:1 with 2 to 4 dB of gain reduction on the loudest lines, a shelf or high pass at 80 Hz to clear rumble, then a limiter on the bus with the ceiling at minus 1 dBTP. That last number is not decoration. AAC encoding can overshoot a digital zero by a decibel, and intersample peaks are what produce that spitty distortion on a cheap laptop speaker.
Read the integrated value only at the very end, over the finished timeline. Reading it over a 40 second section tells you nothing useful.
The targets worth writing on a sticky note
| Destination | Integrated target | True peak ceiling | Quiet files raised? |
|---|---|---|---|
| YouTube | minus 14 LUFS | minus 1 dBTP | No |
| Instagram and TikTok | minus 14 LUFS | minus 1 dBTP | No |
| Apple Podcasts | minus 16 LUFS | minus 1 dBTP | Yes |
| Spotify video and audio | minus 14 LUFS | minus 1 dBTP | Yes, when normalization is on |
| Broadcast delivery, EBU R128 | minus 23 LUFS | minus 1 dBTP | Fixed spec, no offset |
Loudness range, the LRA figure on the same meter, is the one most people ignore. For a talking head video I expect 4 to 8 LU. A documentary with scored sequences and quiet verité sections will read 10 to 14, and that is correct rather than a fault. If an interview film comes in under 3 LU, something in the chain is crushing it flat.
Why minus 14 LUFS is bad advice as usually given
Here is my complaint with the number everyone repeats. Told to hit minus 14, most people reach for the limiter and pull the threshold down until the meter agrees. That hits the target and sacrifices the only thing the target was protecting.
Since the platform merely applies a gain offset, there is no prize for turning up early. A mix that reads minus 16 integrated gets a smaller reduction and plays at a very similar perceived level, with its dynamics intact. Chasing minus 14 by limiting trades headroom for nothing.
The reverse error is the one I made for years. I came out of radio and broadcast habits, so I delivered everything near minus 23 LUFS. On web platforms that is a real problem, because YouTube will not raise you up to meet anyone. Viewers reached for the volume on my films and then got blasted by the next autoplay. I have watched that cost a client about a third of their average view duration on one upload, measured against an identical cut remixed four decibels louder.
So the working rule I give people is a range rather than a value. Land the integrated reading between minus 16 and minus 13 for anything going to the web, do it by gain staging and dialogue balance, and accept whatever offset the platform applies. The music under your intro is a separate job, and the settings I use for it are in the piece on music ducking under speech.
The three checks before you export
Listen on a phone at half volume, held at the distance a real viewer holds it. Words live between roughly 300 Hz and 3 kHz, and a phone speaker reproduces almost nothing below 400 Hz, so a mix that leans on low end warmth for presence falls apart there. If you can follow the speech on a phone, the mix will survive anywhere.
Then play three spots only: the first 30 seconds, the loudest moment, and the quietest moment. Most loudness complaints come from the jump between two of those three, not from the average.
Finally, write down the three numbers with the project. On that interview film the final delivery read minus 16.2 LUFS integrated, 6.4 LU range, minus 1.2 dBTP true peak, and YouTube applied minus 0.9 dB. Having those four figures in my notes meant that when the client asked why episode two felt quieter, I could compare instead of guess. It turned out episode two had a music bed sitting 4 dB hotter under the cold open.
Set the meter up on your master bus template this week and save it as a preset so it opens with every project. Once the loudness side is settled, the codec and bitrate choices in YouTube export settings are what decide whether the picture survives the upload too. More workflow notes sit under video editing.
FAQ
What LUFS should I export a YouTube video at?
Somewhere between minus 16 and minus 13 integrated, with true peak at minus 1 dBTP. Getting there by balancing dialogue against music matters more than hitting an exact figure.
Does YouTube turn quiet videos up?
No. It only reduces files that are louder than its target, so a mix delivered at minus 23 LUFS will simply play quietly against everything around it.
Should I put a limiter on the master bus?
Yes, as a safety net with the ceiling at minus 1 dBTP and only a decibel or two of reduction on peaks. Use it to catch accidents rather than to raise the overall level.
Why does my video sound quieter than others at the same LUFS?
Usually because the competing mix carries more energy in the speech range while yours has loudness tied up in bass or music. Loudness meters average frequencies, so two files can match on paper and feel different on small speakers.
More in Video Editing
Video Editing
How to Build a Paper Edit From Interview Transcripts
Six hours of interviews came back from a shoot in Trieste and I had four weeks to turn them into a 24 minute film.
Video Editing
Relinking Offline Media Without Rebuilding Your Edit
You open a project you were cutting last week and the timeline is a wall of red frames with "Media Offline" stamped across...
Video Editing
Rough Cut vs Fine Cut and What Each Edit Pass Should Fix
Most edits that drag on for weeks do not stall because the footage is bad.