The context
Streaming platforms normalize loudness: they measure each track and adjust its playback gain toward a reference level. They also encode audio in lossy formats (AAC, Ogg Vorbis, Opus) as well as lossless ones (FLAC, ALAC) where available. The master has to work after both.
Platform targets (published or commonly reported values)
| Platform | Reference | Notes |
|---|---|---|
| Spotify | −14 LUFS integrated (Normal); Loud −11, Quiet −19 | also raises quieter tracks, keeping 1 dB of true peak margin; Loud mode applies a limiter |
| Apple Music | about −16 LUFS (Sound Check) | |
| YouTube | about −14 LUFS | generally turns down louder tracks, doesn't raise quieter ones |
| Amazon Music | about −14 LUFS | |
| Tidal | about −14 LUFS | album normalization |
| Deezer | about −15 LUFS |
Spotify recommends keeping true peak below −1 dBTP, and below −2 dBTP if the master is louder than −14 LUFS integrated, to avoid distortion in lossy encoding. Album mode preserves the level differences between the tracks of a record; in playlists and shuffle each track is normalized individually.
What this means for mastering
- You don't need to master at −14: it's the point where the platform will bring the track, not an obligation. Many modern genres are mastered between −11 and −7 LUFS for reasons of density and character; they will be turned down.
- Extra volume gives no advantage in playlists: a master at −7 gets turned down by 7 dB on Spotify. What counts is how it sounds afterward: transients, density, balance.
- A master that is too quiet can be at a disadvantage on platforms that raise little or nothing (YouTube) and on Spotify if the true peak leaves no margin to raise it.
- True peak and codecs: lossy encoding can create peaks above the original file; a margin of −1/−2 dBTP avoids decoder clipping. Listen to an encoded preview (many limiters and Ozone have an AAC/MP3 preview).
- Album consistency: decide the relative levels between tracks with album mode in mind.
Checking tools
- Loudness meters: Youlean Loudness Meter (free), iZotope Insight, Waves WLM, Ozone's meter.
- Sites and plugins that simulate normalization (for example MeterPlugs' “Loudness Penalty” calculator): they show how much each platform will turn the track down.
- Codec preview: Fraunhofer Pro-Codec, Apple Digital Masters Droplet/afconvert, Ozone Codec Preview.
Recommended workflow
- Master for the best sound at the loudness the genre requires.
- Measure integrated LUFS, maximum short-term, true peak, PLR.
- Normalize a copy to −14 LUFS and compare it with normalized references: that's how it will be heard.
- Check the codec preview at −1 dBTP.
- Export at 24 bit (or as the distributor requires).
Example in A minor
Dance-pop single at 122 BPM. Master A: −8 LUFS, true peak −0.3 dBTP: on Spotify it is turned down by 6 dB and the Ogg encoding shows peaks above 0: at risk. Master B: −9.5 LUFS, true peak −1.5 dBTP, PLR 9 dB. Comparison normalized to −14: B has a livelier kick and vocal and no encoding artifacts. B is delivered.
Common mistakes
- Ignoring true peak (especially with very loud masters).
- Comparing with references without normalizing.
- Making a different master for each platform: unnecessary; one well-made master works on all of them.
Exercise
Measure 5 reference tracks in your genre (integrated LUFS, true peak, PLR) and compare them with your latest masters. Then listen to everything normalized to −14 LUFS: where does normalization put you?
Links
“Loudness, LUFS, and Normalization”, “Loudness War and Perceived Dynamics”, “Limiting, Clipping, and True Peak”, “Formats and Delivery”, “Mixing for Social Media and Smartphones”.