LUFS -14, is it a mandatory rule?

[Compyfox]

Bear with me, the following will be a long one, but hopefully it clears a lot of questions.

Basically I didn't understand why the songs of these artists sound louder (surpassing the LUFS). If I adapt my theme to -14 LUFS, when I export the song to WAV and open it with an editor (for example Audacity) I can see that the wave is very small compared with other songs.

I've looked at your waveforms, and the explanation is simple: you didn't "loudness normalize".

Example:
Your track is at -14 LUFS ILk, the track by Power Glove at -9 LUFS ILk, Carpenter Brut is at an insane(!!!, I don't say this lightly) -3 LUFS ILk.

If you just compare the wave forms without touching the gain knob of the channel or the clip gain directly (which also changes the waveform readout most of the time), then of course all other entries will "look" and sound more louder and more impactful. If you pull every track down to -14 LUFS ILk however, then things look (and sound!) a whole lot different. All tracks might now sound "equally" loud - but the other productions lost all their punch due to lack of transients, resulting in a more dull sound and sometimes perceived "less loud" feeling.

That is the main topic of Loudness Normalization (pulling the loudness down) and why it is important to not "master too loud" / retain the transiente.

I did not understand why they ask me to establish -14 LUFS when the artists do not do it and I had the doubt of what I had to do, what is the best (in my case).

My goal in mind, when I was mastering my tracks was that if (let's suppose an ideal situation hehe) in Spotify existed a Synthwave playlist, I'm listening to several artists of this style and, in the middle of the playlist, my song appears, it seemed one more in the playlist. I know that it is impossible in my case (I would need better knowledge and better professional tools ... or maybe to pay to a sound engineer), but at least improve the original song a little bit.

The easiest answer is: because these artist and/or labels can get away with it. Maybe even brand it as "desired" (which is utter nonsense and just a bad excuse if you ask me).


In case of Carpenter Brut, this pushed/distorted type of sound is indeed part of this "desired effect", but it is actually hurting the production in the long run.

As mentioned above, on most platform such loud tracks will be pulled down on the fly prior to playback. Then you can clearly hear the biggest issue: a lack of transients, and the whole thing sounding just like utter mush.

I invested some time to take a deep dive with Carpenter Brut especially. Here is how it looks like on Youtube, what you can listen to right now (I've used Orban Loudness Meter for measuring in this case). Important information off hand: I'm writing LKFS and LUFS. They are the same, only that LKFS clearly tells you "I've used the ITU-R BS.1770-x specs for measurements", while LUFS usually means EBU R-128.







The first image is a screenshot of "Turbo Killer" being measured directly out of Youtube. I did not mess with the PC audio, I did not lower anything on the player itself. As you can clearly see, Youtube loudness normalizes the material and pulls it down by about 11dB, resulting in roughly -15,5 LKFS ILk and a maximum signal strength of about -11dBTP.

The second screenshot is the video itself (outside of Youtube). As you can see, it is not loudness normalized at all. it registers with about -3,2 LKFS ILk and clips the signal into oblivion up to +3,2dB with this measurement tool. Ouch.

From A/B listening... the "normalized" version sounds significantly more quiet, true. But also annoyingly distorted. The "regular super f'n loud release" is actually quite ear fatiguing. I always have to turn off this version after like 1:30min in.




I also took a closer look at the releases through Bandcamp. in fact, "Turbo Killer" was released twice. Once on "Trilogy" and once on "EP III". Both versions are about the same, +-0,2LU offset. Here is a screenshot from one of these versions, measured with Youlean Loudness Meter 2 (thanks to it's nifty drag and drop function - that sadly does not work in the Free version, only the Pro one).





That production is just squashed to utter bits. -3,5 LKFS ILk, barely any dynamic movement, clipping up to +2,7dBTP (different measurement tool, different dBTP readouts). It's as if you're constantly pushed on to a gas pedal. That was the intention probably. And yes, it does work - but not for long.




So what would actually happen if I pull the track down by 6dB?







Here is a comparison between a Izotope RX7 de-clipped version (first screenshot), and just a "gain reduced" one (second screenshot). I do have to point out that even the "de-clipping" didn't help - the track is sadly beyond repair.




"Can 'we' do better?!"

Funny enough: yes, we can!

The following are two screenshots from the "Music Video" version of "Turbo Killer". Although please note that the first screenshot of the Youtube video shows an offset of about 0,3LU, because I created that one in the middle of the song.







Important to point out here, is that this is a music video! So different rules apply than for a regular music release. However, if you take a closer look at the waveform, you will notice that this song has transients (and by that, I don't mean the strong peaks from the sound FX) - something that is gone from the -3,5 LUFS ILk release. So the music for this particular video used a more reasonable loud final mix. And funny enough, this track is suddenly less ear fatiguing and feels less annoying. In fact, the organ is way more haunting, the transition between organ and pulsing bass/beat is way more exciting, you actually do not want this video to end. IMHO and all that.





Let's summarize:

• if you feel that your track is "too quiet" for a regular release - that is a valid concern, but as this short analysis clearly shows - louder is not always better. It only offers you a temporary impact, but can result in ear fatigue really quick
• in the long run, more and more streaming services adapt "loudness normalization on playback" - loud tracks will be pulled down. Sometimes way too quiet tracks will be "pulled up" to have a more evened out broadcast (especially the case on radio and TV, also Spotify Loud).
• creating multiple mastered versions for several platforms is a lot of extra work and a huge pain. Setting one global "suitable" loudness (e.g. -14 LUFS ILk absolute max) automatically solves a lot of problems and offers way more release possibilities
• too loud masters will suffer the most due to a lack of transients. People want to buy an album like they've heard on a loudness normalized streaming service - if they then listen to it "outside" of a loudness normalized environment, they will drop off their chairs. Selling different loud versions will only result in frustration (I've had this discussion years ago with plenty of re-releases of "Beatles" albums, but also with Metallica's "Death Magnetic" official CD release vs the Guitar Hero version)

So to get back to your main question: "LUFS -14, is it a mandatory rule?"

A lot of people will say "not really". But the reality is, that it will be an upper limit in the foreseeable future (2-4 years). Your focus should be on creating a good mix that can work and is enjoyable at any given volume. If you stay below -14 LUFS ILk (ideally even -16 LUFS ILk) after your finalized mix (meaning: properly mixed, "fairy dust" pre-mastering - not "setting release loudness levels"!), then you've basically evaded the Loudness War - and your listeners will thank you for it.




A lot of stuff to get through, but I hope this answers your question. Maybe even for those that were afraid to ask so far. Have a nice evening!

[jcea]

Good day!, Sorry my delay, it took me a while to read all the information you gave me .
I sincerely thank you very much for even bothering to analyze the Carpenter Brut song . I hadn't thought of comparing the official song with the music video and you are right, I notice the music video version is more dynamic (even the sound of the main lead sounds a bit more varied to me, I don't know if it is an effect of the mix itself or that it is really its real sound).

I've looked at your waveforms, and the explanation is simple: you didn't "loudness normalize".

Sorry to ask you again , but I don't know if what you want to tell me is that I didn't "loudness normalize", and that's fine, or I didn't "loudness normalize", and I should because that's wrong.

Otherwise, in general I have understood everything you have written. Personally I also think that mastering is not to make everything sound louder and that all the parts of your song sound louder, in fact I waste quite a lot of time (once I have the song finished) in mixing each channel well so that each sound sounds different (maybe I'm interested in making an arpeggio sound a little quiet so that it does not "hide" the main melody, for example).

I will continue practicing and studying more about this.

What I am very clear about is that the world of mastering has become another "fun" task for me (besides creating music) and I hope to get some practice and improve my songs a little bit .

I hope to share my songs with you, unmastered and mastered, to know your opinions/recommendations .

By the way, I understand that there is also a standardization for panning, personally I love (and may abuse a little) to use panning in my songs (going back to the example of the arpeggio, the arpeggio could "travels" from the left to the right channel) but I am not taking into account (nor have I read yet) those standardizations. I will do that later, in my future self, when I control a little more the mastering hehe .

Thanks again to all of you for your contributions, I think the thread has been very interesting and very instructive (at least for me hehe).

And thank you very much for your kindness, because of my inexperience, I was a little afraid to ask and say something stupid .

We keep in touch in the forum .

Thanks again guys, cheers!

[excuse me please]

Thanks for the info! The Orban has a free version FYI, just downloaded it

[Kwurqx]

Apart from the LUFS / Dynamic Range discussion, which is mostly about the integrated "averaged" values, there is still the issue of the physical upper limits. So, about the upper limit: there is a max above which distortion will happen. Especially the hard maximum/ceiling of the digital 0 dBFS.

Since you can't go over 0 dBFS (the signal peaks will be clipped), you should minimize the chance of ever reaching it. To complicate things, due to samplerate/resolution and audio encode/decode inaccuracies, (inter)sample peaks may occur in reproduction, where they do not occur in the original source material. Generally some headroom (at least some -0.3 dBFS) is recommended to allow for these "inaccuracy" peaks.

[Compyfox]

Sorry to ask you again , but I don't know if what you want to tell me is that I didn't "loudness normalize", and that's fine, or I didn't "loudness normalize", and I should because that's wrong.

For clarification:
You wanted to compare music by Power Glove, Carpenter Brut, etc to your own content.

Your content is at -14 LUFS ILk
Power Glove (I've checked a couple of entries, most notably "Playback" and "Secrets" from Bandcamp) is about -8,3 LUFS ILk
Carpenter Brut ("Turbo Killer") is at unholy -3,5LUFS

So you're comparing tracks as they are "released", not in any objective manner. To compare tracks objectively, you need to pull the commercial releases down to the same loudness that your track is. Or pull up/down all tracks to a certain desired target LUFS value. Only then you can really hear possible issues, or find out what your production might still miss.


That means:
Let's say the desired Loudenss Target is -14 LUFS

Your content remains at -14 LUFS ILk
Power Glove - "Playback" needs to be pulled down by 5,7dB to 6dB to get from -8,3 LUFS ILk to -14 LUFS ILk
Power Glove - "Secrets" needs to be pulled down by 2,7dB to 3dB to get from -11,3 LUFS ILk to -14 LUFS ILk
Carpenter Brut - "Turbo Killer" needs to be pulled down by 10,5dB to 11dB to get from -3,5 LUFS ILk to -14 LUFS ILk (and that doesn't work in this tracks favor at all!)


That is what I meant with "Loudness Normalizing". You don't "pull up" (raise the loudness), you pull everything else down to you. You only adjust up/down if you want to compare at a different LUFS value.

Thanks for the info! The Orban has a free version FYI, just downloaded it

The PC/Mac software version of the Orban Loudness Meter has been free for a couple of years at this point. It was sadly never updated to utilize ITU-R BS.1770-4 (current version), it only goes up until ITU-R BS.1770-2.

Still great for quick checkups, especially as offline tool.

Apart from the LUFS / Dynamic Range discussion, which is mostly about the integrated "averaged" values, there is still the issue of the physical upper limits. So, about the upper limit: there is a max above which distortion will happen. Especially the hard maximum/ceiling of the digital 0 dBFS.

Since you can't go over 0 dBFS (the signal peaks will be clipped), you should minimize the chance of ever reaching it. To complicate things, due to samplerate/resolution and audio encode/decode inaccuracies, (inter)sample peaks may occur in reproduction, where they do not occur in the original source material. Generally some headroom (at least some -0.3 dBFS) is recommended to allow for these "inaccuracy" peaks.

Actually, an even bigger headroom is recommended. Especially the stronger you push loudness, and convert to MP3 on top. Sadly not a lot of people adapted this (back then), and I still see limiters with presets that have a ceiling of -0,1dBFS. Even back in the day at -8dB RMS avg, I used -0,5dBFS as headroom, because I had equipment (for testing purposes) that started to barf if any signal went higher in terms of peaks. Push to a "requested -4,5dB RMS avg", and that headroom actually needs to be -1dBFS.

This is why the LUFS metering has a set maximum of -1,0 dBTP, France, Japan and certain US areas even have -2,0 dBTP as safety mechanism during broadcasts/streaming.

[jcea]

Thank you very much again Compyfox for bothering to give me such detailed answers, it's a pleasure to read you and I didn't expect you to analyze the artists I mentioned, it helps me to understand better what you are telling me .

Cheers!

[FarleyCZ]

Lemme join the discussion real quick. I think it's good that volume normalisation is becoming standardised, but I wouldn't boil it down to a sipmle "here's your target value" kind of advice. See, this thing has been happening for years in broadcast already, where EBU released the R-128 recommendation years ago. Not only very few people understood it in the industry (cause you need to know what the hell LUfs is and how it's different from a normal dBfs, which is described in completely different norm (ITU.1770), but it's been a hot discussion topic ever since.

The problem is the integration. That works wonders if you have a material with fairly consistent levels. So as long as there's a talking voice after a talking voice on the TV, R-128 compliance pretty much destroys the loudnes wars. But it breaks down, when you have a material with different distribution of quiet passages. Like a football match with small amount of spectators on the tribune, or a classical music concert with a huge dynamic range.

Same goes with "14 LKFs" rule. As long as those songs are similar in length and have statistically similar amount of break downs and quiet parts, it can heal the loudess war. But once you have anything out of the norms (for example I like a really long breakdowns in otherwise quite dynamic music), the rule is not relevant.

It just isn't that easy.

[Compyfox]

That is a true and valid argument that is continuously debated a lot, especially behind closed doors. The main difference between "Broadcast Stations", "Streaming Platforms" and just regular stores, is (indeed) how the integration is handled.

Broadcast Stations use a certain technical array to never stray too far from the parameters/targets they have to stay in. In case of Streaming Platforms, the biggest debate is still "Integrated Loudness per whole album, or per individual track?!". And stores just don't have this yet.


It is true, that an album should have an overall integrated loudness. So that you can listen to the whole thing from start to finish, have the full dynamic experience and never really touch the volume dial. It is "one concept" after all. This is especially apparent with radio plays, concept albums, soundtracks, etc. This takes some getting used to in terms of editing, but these days it's easily possible to pull off.

Most streaming platforms are going by random individual tracks (think shuffle mode) however, and here things are definitely different. Although I remember debates even back in 2016, that there needs to be a second meta parameter that declares "part of an album, this LUFS is correct - don't adjust" (which was, again, publicly addressed at AES142 in Berlin, Germany - I was sitting in that panel). But we're sadly still far from implementing this - over three years later!

I think my last status on that was, that the engines are/were supposed to look at the album LUFS first, then the individual track LUFS -- and only then adjust accordingly.

Example:
Album LUFS actually being -12 LUFS ILk, the most dynamic/quiet track you want to listen to (the ballad of the album) being maybe -15 LUFS ILk. The streaming service target is -14 LUFS ILk... So while listening to a -14 LUFS ILk stream, that particular track is being pulled down by 2 LU to -17LUFS. But it won't feel out of place in the rest of the stream - because it still has it's "loud parts" which are still comfortable to listen to (and blend quite nicely with the rest of the feed) since it was part of a whole concept.




I hope that most of us can agree on the following:

The IMO most strict "rule" we can (and ideally should) set up here, is to "never exceed a certain LUFS value". For the sake of having better enjoyable productions, retaining dynamics / transients. And I think in this case, -14 LUFS ILk +-0LU (as in - NEVER EVER go higher) is a good target IMO. Although I already hear the arguments from far away - then maybe use -12 LUFS ILk +-0LU as absolute max for individual music tracks until August 2021 (nine more months!) - and that's it. After that, no more debates.

I think we've had a long enough transition period - starting with the K-System (v1) in early/mid 2000s. That is over 15 years! Most streaming services adapted -14 LUFS at this point. A deadline worked for the Broadcast realm, it also worked for light bulbs. Why not also for music?

Let's stop pretending that "certain tracks/genres (still) need this" - because you've been lazy regarding your overall sound design and rather slam everything through a master limiter array until all Christmas trees lit up.



That is my personal take/opinion on the "Loudness War" however. If this is the only way to force people to learn to do it "in a better way" again... I'm all for it.

[jcea]

Hello again!

Well, after all I've read, I've tried to do some mastering to 2 songs. I would like to know your honest opinion (at least, if it sounds better than the original mix xD), I'll pass you the link:

https://drive.google.com/drive/folders/ ... sp=sharing

Basically:
- Experiment_7/8.wav: Original theme, unmastered (so you can see the original mix)
- Experiment_7/8_New_Mastered.wav: Mastered theme, without following LUFS recommendations (I mean, that I thought it sounds good)
- Experiment_7/8_14LUFS: Same theme mastered, lowering a little the limiter so that it arrived at -14LUFS, to fulfill the recommendations.

The mastering chain I used more or less was the following:
- Equalizer to eliminate unnecessary frequencies: 35hz (backwards) in bass and 19.5Khz (forward) in treble
- Mid/Side Equalizer to centralize the bass (the drums mostly, the bass being synthesized I think it is better not to put it in)
- A compressor but compressing very little, I still don't control the compression issue very well and I preferred not to risk too much so as not to spoil the song
- Multiband compressor (OTT plugin). The truth is that it seemed to me more like an equalization stage than a compression one, maybe I have to learn how to handle it better xD.
- The suite Ozone 9 to put him a touch of final equalization and, finally, a limiter (Maximizer) to adjust the LUFS
- The Izotope Tonal Balance Control, to see that the bass, midrange and treble are more or less in place.
- Finally, the Spectrum Analyzer (to see that the frequencies are more or less stable) and the Youlean to see the LUFS.

Let's see what you think . Of course, any comment/criticism will be very well received and I will be very grateful

Cheers and thanks!!