DUCKER

Mastering to a loudness target

Streaming platforms turn everything down to the same loudness, which changed what mastering is for. Here is what the numbers mean and which one to aim at.

LUFS, in one paragraph

LUFS — Loudness Units relative to Full Scale — is a measure of how loud something sounds, not how big its waveform is. It weights frequencies the way human hearing does, so a bright cymbal counts for more than a sub-bass note of the same amplitude, and it averages over time. One LUFS is one decibel, so the scale reads like dB and the numbers are negative because they are measured down from digital full scale. LKFS is the same measurement under a different name, used in American broadcast.

Three readings matter: integrated (the whole piece, which is the number platforms use), short-term (the last three seconds) and momentary (the last four hundred milliseconds). When someone gives one number, they mean integrated.

What each platform normalises to

These are not mastering targets so much as what happens to your file on the way to a listener. Anything louder is turned down; some services turn quiet material up as well.

Where it is goingNormalised toTrue peak
Spotify−14 LUFS−1 dBTP
Apple Music−16 LUFS−1 dBTP
YouTube / YouTube Music−14 LUFS−1 dBTP
Amazon Music, Tidal−14 LUFS−1 dBTP
Apple Podcasts−16 LUFS ±1−1 dBTP
Podcast, mono file−19 LUFS−1 dBTP
Broadcast television (EBU R128)−23 LUFS−1 dBTP
Broadcast television (ATSC A/85, US)−24 LKFS−2 dBTP

A mono podcast really does need a different number. A mono file played through two speakers measures quieter than a stereo file at the same reading, so −19 LUFS mono lands at about the same perceived loudness as −16 LUFS stereo. Several tools do not correct for this, which is why some mono episodes arrive noticeably quiet.

Why "make it as loud as possible" stopped working

Before normalisation, a louder master genuinely won: it sounded bigger next to a quieter one on the radio or in a playlist. That is what drove thirty years of squashing everything flat. Now, if you deliver at −8 LUFS to a service that normalises to −14, the service simply turns your file down by six decibels. The loudness is gone and all the compression you used to get it is still there — so the master is quieter and flatter than one that was left alone. Dynamic range is the only thing that survives the trip.

The practical consequence: master for the sound, land somewhere near the target, and keep the peaks under control. Do not chase the number.

Peak and true peak are not the same

A digital file's sample peak is the highest sample in it. But a converter, or an MP3 or AAC encoder, reconstructs a smooth curve between the samples, and that curve can go higher than any sample did. True peak (dBTP) measures that reconstructed curve. A master that peaks at exactly 0 dBFS can easily hit +1 dBTP after encoding, which is distortion in the file every listener receives. That is what the −1 dBTP ceiling is for: it is headroom for the encoder, not a suggestion.

An order of operations that works

  1. Fix the mix first. Mastering cannot rescue a mix where two elements fight. If you have the stems, that is where the problem is cheaper to fix.
  2. Balance the broad tone. Gentle, wide moves — a decibel or two of low end, a decibel of air. If you need a 6 dB cut, go back to the mix.
  3. Control dynamics, lightly. Glue, not squash. If the meter is moving more than a few decibels, you are mixing, not mastering.
  4. Limit to the true peak ceiling. −1 dBTP for anything that will be encoded.
  5. Measure integrated loudness and set the level. Not by ear against a commercial reference at a different loudness — by the meter.

Doing it in the browser

Ducker's Master page takes your dialogue, music, effects and ambience as separate stems, ducks and carves them against each other, measures integrated loudness on the finished render, and applies the makeup gain to reach the target you chose — −23, −14 or −11 LUFS — with a limiter holding the true peak down. It renders offline rather than in real time, so a five-minute mix takes seconds rather than five minutes, and it exports lossless WAV or MP3, as one mix or as separate mastered stems.

Files are decoded in your browser. Nothing is uploaded, and there is no account to make.

Open Ducker and master a mix → Free, in the browser. Drop in stems, pick a target, export.

Common questions

What LUFS should I master to?

−14 LUFS integrated covers Spotify, YouTube, Amazon and Tidal. Use −16 for Apple Music and for a stereo podcast, −19 for a mono podcast, and −23 LUFS or −24 LKFS only when a broadcaster asks for it. Keep true peak at −1 dBTP in every case.

Is −14 LUFS a rule I have to hit exactly?

No. It is where playback lands, not a specification. Within a decibel or two is fine, and a quieter, more dynamic master is usually better than one squashed to hit the number precisely.

What is the difference between LUFS and LKFS?

Nothing measurable. LKFS is the ITU and ATSC name, LUFS is the EBU name, and both describe the same K-weighted loudness measurement. American broadcast documents tend to say LKFS.

Why does my master sound quieter than commercial releases?

Usually because you are comparing at different loudness. Match levels first. If it still sounds smaller, the difference is arrangement and mix balance, not the limiter.

Can I master audio for free online without uploading my files?

Yes. Ducker processes audio locally in the browser using the Web Audio API, so your files never leave the machine. Most online mastering services upload to a server; this one has no server to upload to.