Guide
Two numbers, and only one of them is loudness
Aim for -14 LUFS integrated and hold your true peaks at -1 dBTP. The first number decides whether you sound as loud as the feed around you, the second decides whether you sound clean, and mixing them up is the most common audio mistake in short-form video.
- LUFS integrated, the working target
- -14
- dBTP ceiling, measured after the encode
- -1
- Passes: measure first, then apply
- 2
Integrated loudness is an average, true peak is a ceiling, and they fail differently
LUFS integrated is one number for the whole file. It is a loudness average weighted the way human hearing works, so a quiet passage and a shout average out into a single figure.
dBTP is not an average. It is the highest point the waveform reaches when a player reconstructs it, and it is the number that decides whether anything crackles.
You can be quiet and clipped at the same time. A read with three loud consonants and a lot of silence measures -22 LUFS and still slams into the ceiling on every plosive.
That combination is the one people misdiagnose. They hear distortion, decide the file is too loud, pull the fader down, and get a file that is both quiet and still distorted.
LRA is the third number and almost nobody sets it. It caps how far the loud and quiet parts are allowed to sit apart. We hold it at 11, which is wide enough to keep a read sounding like a person and tight enough that nobody reaches for their volume.
The two numbers and what each one is telling you
Both can fail on the same file, and they fail in opposite directions, which is why one fader cannot fix both.
| Integrated LUFS | True peak dBTP | |
|---|---|---|
| What it measures | Average loudness, whole file | The single highest point |
| Symptom when wrong | Sounds weak next to the feed | Crackle, spit, harsh esses |
| Where you hear it | Scrolling past other videos | Phone speaker, cheap earbuds |
| The fix | Gain the whole file | Limit the peaks only |
No platform publishes its curve, so -14 is a landing zone rather than a spec
Be honest about what this target is. TikTok, Instagram and YouTube all apply their own loudness handling and none of them publishes the numbers.
Published advice in the wild spans about -16 to -10 LUFS, and the people giving it are all guessing from measurements of finished files.
So -14 is not a rule anyone can enforce. It is the middle of the range everyone else lands in, which is the whole point: you are not trying to hit a spec, you are trying not to be six decibels below the video before yours.
The difference between -14 and -13 is inaudible. The difference between -14 and -22 is somebody reaching for the volume, and half of them do not bother.
This is why the ceiling deserves more of your attention than the loudness. A file at -16 sounds fine. A file with clipped peaks sounds broken on every phone speaker it touches.
Where the common advice sits, in LUFS
The spread is real and nobody can settle it, because the platforms do not say. Anywhere in the middle is fine; the bottom of the range is not.
- Loud short-form advice-10 LUFS
- What we deliver-14 LUFS
- Conservative advice-16 LUFS
- An un-normalized phone recording-24 LUFS
This is the whole editor
Highlight a phrase and a clip lands on those exact words. No timeline, no keyframes, no layers.
Measure the file first, then apply the gain, or you will pump the mix
A one-pass normalizer guesses at the gain while the file is playing, which means it moves the gain during the file. If there is music ducking under a voice, that movement fights the ducking and you hear it breathe.
Two passes fix it. The first pass measures and writes nothing. The second applies one uniform gain across the whole file, so every level relationship you built survives intact.
The order matters more than the tool. Any editor that reports integrated loudness can do this: measure, work out the offset, apply it once.
Then limit, not before. A limiter in front of the gain stage is guessing at a level that has not been set yet.
The order that keeps a mix intact
Every step is reversible except the last. Doing them out of order is what makes normalized audio sound squashed.
Measure
Analysis only, nothing written
Apply one gain
Uniform, so ducking survives
Limit the peaks
Under the target, not at it
Encode
The encode adds peak of its own
What hitting -14 costs you, said plainly
Normalizing upward raises everything, including the room. A quiet read in a kitchen with a fridge running arrives at -14 with an audible fridge.
There is no processing answer to that. Get the microphone closer, turn the fridge off, and the same normalization that exposed the noise stops being a problem.
Loudness normalization also cannot rescue a recording that was already clipped at the source. Once the peaks are square, the information is gone and every later stage is polishing damage.
And if music is the point of your video rather than a bed under a voice, this whole target is the wrong one. A track meant to be listened to gets mastered for listening, not for a feed.
Questions people ask
- Is -14 LUFS an official TikTok or Instagram requirement?
- No. Neither platform publishes a delivery loudness target, and anyone quoting one as official is inferring it from measurements. Treat -14 as the middle of the range other people land in rather than a spec you can be marked against.
- Why -1 dBTP and not 0?
- Because 0 leaves no room for the encoder. Converting to AAC adds true peak that was not in your file, so a master sitting exactly at 0 comes out the other side above it. The headroom is what absorbs that.
- My editor shows dB, not LUFS. Can I still hit this?
- Not reliably. A dB meter shows peak or RMS, and neither maps onto integrated loudness. Most modern editors have a loudness meter hidden in the audio workspace, and free standalone analysers will read a finished file in seconds.
- Does Cutroom let me choose the target?
- No. Every export normalizes to -14 LUFS with the peaks held under -1 dBTP, and there is no control for it. That is the right default for an ad in a feed and the wrong tool if you need to deliver to a broadcast spec.
Set the loudness once, protect the ceiling always, and stop worrying about the decimal. Every Cutroom export leaves at -14 LUFS with the peaks already held, because it is not a setting worth making anybody learn.