What LUFS should you master to in 2026?
Every major streaming platform turns your track up or down to sit at roughly the same loudness as everything else. This is loudness normalisation, it has been standard for years, and it means the loudness war is over — you just cannot tell from inside your DAW.
The practical consequence surprises people: mastering your track much louder than the platform target does not make it louder for the listener. The platform simply turns it down. What you keep is the damage — the squashed transients and reduced dynamic range you traded away to get there.
The targets, platform by platform
These are integrated LUFS — loudness measured across the whole track, not a peak reading. A track measured at −14 LUFS integrated will be played at about the same perceived level on Spotify as any other track at −14.
| Platform | Normalisation target | What it does if you are louder |
|---|---|---|
| Spotify | −14 LUFS | Turns down. Optional "loud" setting is −11. |
| Apple Music | −16 LUFS | Turns down. Sound Check is on by default. |
| YouTube | −14 LUFS | Turns down. Never turns quiet tracks up. |
| Amazon Music | −14 LUFS | Turns down. |
| TikTok / Instagram | ≈ −14 LUFS | Turns down, and heavily compresses on top. |
YouTube only ever attenuates — it will not raise a quiet track. So a master at −20 LUFS stays quiet there, while on Spotify it may be turned up. Master to the target, not below it.
So what do you actually aim for?
For most music, master to around −14 LUFS integrated with true peak at −1 dBTP. That sits correctly on Spotify, YouTube and Amazon, and Apple will turn it down 2 dB, which costs you nothing audible.
The −1 dBTP ceiling matters more than people expect. Streaming services encode to lossy formats, and lossy encoding can push peaks above where they sat in your WAV. Leaving a dB of headroom is what stops that becoming audible distortion on the listener's phone.
Genre is a real exception. Quiet, dynamic music — classical, ambient, sparse acoustic — should not be pushed to −14. Let it be −18 or −20 and keep its dynamics; the platform will not penalise it, and squashing it to hit a number is the exact mistake normalisation was introduced to make unnecessary.
- Pop, rock, electronic, hip-hop: −14 LUFS integrated, −1 dBTP
- Podcast and spoken word: −16 LUFS (mono −19)
- Classical, ambient, dynamic acoustic: −18 to −20 LUFS, keep the range
- Never exceed −1 dBTP regardless of genre
Loudness is not the same as being finished
A track that measures −14 LUFS can still sound worse than one at −16, because integrated loudness says nothing about tonal balance, phase, or whether the vocal is audible. Loudness is the last decision, not the first.
If a master sounds thin or harsh next to a reference at the same measured loudness, the problem is the mix, and no limiter setting will fix it.
Doing it in dhun
Ask for a loudness target directly — "master this to −14 LUFS for Spotify" — and it uses pyloudnorm to hit the integrated target exactly, with true-peak limiting.
If you want it to sit like a specific record rather than hit a number, upload the reference and ask it to match. That uses Matchering, which matches tonal balance as well as loudness, and it will tell you how far apart the two were.