We’ve all heard about Loudness and LUFS by now. On Production Expert, thanks principally to Mike Thornton’s encyclopaedic coverage, what started as something that post-production mixers needed to worry about has gently transformed the music listening experience of the majority of people and by presenting a workable route out of the decades-long dead-end which was the Loudness War, has moved music production on to a new chapter.

So what’s new? Well, while nearly every major streaming service has been using Loudness Normalisation for years, there was one which wasn’t. Apple Music. That has now changed.

The idea is simple enough, just like a human listener would turn up quieter sounds and turn down loud ones, Loudness Normalisation does exactly that. Because ‘loudness’ is perceptual rather than a directly measurable parameter of the audio file like the peak sample level in old-school Peak Normalisation, a method for reliably measuring perceived loudness needed to be developed, which was more difficult than it sounds. But, once BS1770 and the Loudness Unit (LU) was developed, the mechanism to automatically match the perceived loudness of audio became doable.

What Loudness Level Do Streaming Platforms Use?

Over the last five or so years, Spotify, YouTube, Tidal, Amazon and SoundCloud have all settled on a LUFS normalisation value of -14LUFS. Apple has announced their LUFS target to be significantly quieter at -16LUFS. However, by choosing -16LUFS, Apple has taken on board the recent AES TD1008 recommendations. To learn more about this check out our article AES Update Loudness Recommendations For Audio Only Streaming. For completeness, it should be mentioned that Spotify also offers a -11LUFS Loud option and Deezer Normalise to -15LUFS.

Sound Check is now on by default on iOS devices making loudness normalisation transparent to the consumer.

What Does This Mean For Engineers?

Apple is too big to be ignored and this move brings consistency, but no convergence to a single universal level. The obvious move, considering the current landscape, would have been for Apple to choose -14LUFS but considering this move to -16LUFS and for every extra reduction in LUFS comes with it an accompanying headroom bonus opening up more dynamic range, this is a welcome move.

The fact that the new level is lower is also interesting because it would appear that Apple is taking on board the excellent work done by the Audio Engineering Society. To help understand the issues, we asked Bob Katz and Rob Byers, who are part of the group that drew up the AES recommendations, both the original TD1004 and the revised TD1008. You can either listen to the podcast or read the article which has a transcript of our conversation.

It is gratifying to see Apple take a lead and start the trend to drop the target loudness for streaming services. Let’s hope that other services follow Apple’s lead. However that is very far from inevitable and for the time being remains a hope on my part.

That said, the effect Apple can have on the development of audio streaming is illustrated perfectly in the announcement of Spatial Audio, which accelerated interest in mixing music for Dolby Atmos enormously, even though Spatial Audio isn’t quite the same thing as Dolby Atmos.

Does This Mean I Should Be Mastering For -14 And -16 LUFS?

While products that allow service by service auditioning of what you master will sound like at the normalisation levels of different streaming platforms, to do this exhaustively might be to be missing the point. Mastering engineer and Loudness expert Ian Shepherd proposes an alternative approach which matches my own experience. He suggests concentrating on the Short Term Loudness reading rather than monitoring the Integrated value. He suggests going no louder than -10LUFS in the loudest sections. He also stresses the importance of True Peak levels.

Peak Levels And Dynamics

The significance of peak levels can’t be overstated as loudness is relative but peak levels are absolute. The reason this is as important as is it today is the variety of different encoding algorithms audio is likely to pass through as it propagates through the various streaming platforms. In the old days of lossless PMC audio and CDs, peak levels could be pushed very close to full scale, exactly how close was a point of contention but close nonetheless. In these days of mp3, AAC and other lossy algorithms, depending on the algorithm used and the bitrate chosen, the maximum peak level before distortion varies and between 1 and up to 3dB of clear headroom between your True Peak level and full scale is recommended.

I recently tested Sonnox ListenHub and the layout of the metering of this plugin illustrates this point perfectly. It offers a momentary Loudness reading, but only for Short Term Loudness. It also features a dynamics meter with a scrolling history histogram. The prominence dynamics are given over loudness illustrates the fact that loudness changes with playback volume but the internal dynamics of a track do not. As long as your track’s loudness is in the right ballpark but your dynamics are intact and your True Peak level is appropriate then it shouldn’t suffer a negative impact from the Loudness Normalisation applied by any of the streaming platforms.

How loud your mix is of secondary importance to the internal dynamics of the mix, that’s the whole point.

Photo by Brett Jordan on Unsplash