do streaming platforms apply loudness normalization?

[beagleman]

I have playlists, both local and streamed that mix classical, jazz, and various eras of pop (from Cole Porter to recent).

I can go from something like hard rock to Lark Ascending. The level change is sometimes disconcerting, but I like it.

I love all types of music, well jazz, classical and classic rock...for my ALL types, but.......

Some classical, not sure there is any exact period, maybe the 60s and 70s, but some was just recorded at quite lower levels, even though of course there are possibly some loud parts also.

I have noticed a lot of classical maybe in the last 30 years, avoids that "Super low" level stuff they used to do, I think it was based on fear of saturating the tapes of the time perhaps.
I have to look through my stuff to find examples, but my normal volume control level is maybe between -22 to -16db on most classical, but some I have to boost all the way up to maybe -10db to get a decently loud volume.

 

[danadam]

You either need an album with integrated LUFS above -14

Actually, the way the album normalization works on Spotify (not sure about other streaming platforms) is that the track gain of the loudest track in the album becomes the album gain of that album. So for example if the the whole album is -14 LUFSi but one of the tracks is at -8 LUFSi then the album will be played 6 dB quieter when normalization is enabled.

I have to look through my stuff to find examples, but my normal volume control level is maybe between -22 to -16db on most classical, but some I have to boost all the way up to maybe -10db to get a decently loud volume.

For me, that "Guardian Angel" mentioned earlier qualifies. The version I bought in files is additionally 6 dB quieter than what's on Spotify. The album is at -34.3 LUFSi and the peak is at -13.34 dBFS.

 

[RegularStereoGuy]

Think about that next time you change channel on TV and make sure you have responsive volume control's trigger happy.
Yes it is made purposely for broadcastings, it pretty much works for all, catch is you need to have attached metadata or encode in it before streaming. Exactly what we aren't getting.

Which brings up a point I made, regarding analog vs digital audio in general on Gearspace and other such forums:

I preferred the analog era, with it's VU or other such meters where Zero was between 2/3 and 3/4 of the way up the scale, rather than at the top, as with digital dBfs.

I believe that the latter situation has led to the phenomenon which I call "chasing zero", which I suspect has partly contributed to the trend toward louder and more compressed music releases. On Gearspace, my theory nearly got me "chased" off the boards.


With analog VU, there is less, not none, but less loudness variation between pieces mixed and mastered, on the same chain.

@beagleman no matter how good EBU 128 is made it's not magic and fun is fun and done is done. What's ruined with over gate compressor and down to 4~5 DR can't really be fixed just not shouting loud. EBU doesn't change DR, doesn't do gate compression/expansion and it's not peak based limiter. It does it in sequences (peace's) to them self, each other and to absolute scale as outputted (per and between chenels). Spotify disassembled it, put their own LUFS target's to three lv they classified instead automatically independent based on segments as it is in EBU as that needs much less metadata attached obviously. So did many other including Apple. Their argument are various from that they don't need -23 dB as output lv or how for music that's too much or how mobile devices don't have enough power to put output that low. And they are all true but it doesn't fix the problem and you will have substantial SPL lv output difference between them, other streaming and broadcasting services so you end up at the beginning of the problem. Of course you can bust LUFS a lot when you do it in floating point precision and glue it back to integer not introducing noise so up to -23 dB (it's not -23 more like -11~-12) can be compensated back and still stay in line with everything else (absolute level's).

When they lower LUFS threshold the materials out of range (with higher DR than -LUFS) it doesn't get processed as it's out of reach as you found out your self with very high (16~17) DR classical music and not adopted THX cinema movie's can go up to 24 DR.
Edit: Wiki article is actually decent all do it doesn't go down in details.

EBU R 128 - Wikipedia

en.m.wikipedia.org

I would agree to a compromise across the board of -20LUFS/dbfs/Whatever this digital thing has made leveling so complicated.

Agreed on all your points that loudness normalization is simply a sophisticated automated volume knob, does not compress or EQ. Depending on the threshold, some normalizers will, in extreeeeme cases apply mild limiting to get a piece to match the target loudness, or, in the case of a hyperdynamic song, normalize to 1-2LUFS below target and preserve those peaks.

If I owned a fully diigital recording or mix setup, I'd still run everything through a final VU meter to get it down to at least pre-1990 levels, and lazy folks would have to raise their volume settings a little higher waah waahh!

 

[ZolaIII]

@RegularStereoGuy it whose and is about chasing to bury the noise below the signal and that's the main difference between digital and analog. I whose one of those guys who chased zeros. Lucky it's a done thing the rendurance of back rounding from floating point to integer. Now noise is not relevant factor on let's say double precision FP, same as loudness normalisation on signal side should be. There always where good, bad and ruined recordings. In old analog time's they were much more limited by noise threshold and lowering the level whose genuine in masking it. Same as too little DR is bad so is too much and then you have SFX that threatening to bring the hause down and speech dialogues you still can't hear. Psy is a key factor but most people in any camp fail to realise it. Perceptually ±12 dB is twice as loud or half as loud, the good old equal loudness compensation (newer implementation as ISO226). Gate compressor when used properly is perfectly fine thing especially for such with huge DR. VU's while looking good to many people, where too slow with too little threshold, even today's measurement small capsule microphones can't catch true peek's accurately or RTA it realtime.

 

[RegularStereoGuy]

@RegularStereoGuy it whose and is about chasing to bury the noise below the signal and that's the main difference between digital and analog. I whose one of those guys who chased zeros. Lucky it's a done thing the rendurance of back rounding from floating point to integer. Now noise is not relevant factor on let's say double precision FP, same as loudness normalisation on signal side should be. There always where good, bad and ruined recordings. In old analog time's they were much more limited by noise threshold and lowering the level whose genuine in masking it. Same as too little DR is bad so is too much and then you have SFX that threatening to bring the hause down and speech dialogues you still can't hear. Psy is a key factor but most people in any camp fail to realise it. Perceptually ±12 dB is twice as loud or half as loud, the good old equal loudness compensation (newer implementation as ISO226). Gate compressor when used properly is perfectly fine thing especially for such with huge DR. VU's while looking good to many people, where too slow with too little threshold, even today's measurement small capsule microphones can't catch true peek's accurately or RTA it realtime.

Well, my engineering 'zero' is -18dBfs on a pure peak-based meter: plenty of headroom if someone wants it loud and squashed later on.

 

[RegularStereoGuy]

I love all types of music, well jazz, classical and classic rock...for my ALL types, but.......

Some classical, not sure there is any exact period, maybe the 60s and 70s, but some was just recorded at quite lower levels, even though of course there are possibly some loud parts also.

I have noticed a lot of classical maybe in the last 30 years, avoids that "Super low" level stuff they used to do, I think it was based on fear of saturating the tapes of the time perhaps.
I have to look through my stuff to find examples, but my normal volume control level is maybe between -22 to -16db on most classical, but some I have to boost all the way up to maybe -10db to get a decently loud volume.

Even concert ('classical') music recordings have succumbed, albeit to a lesser degree than the popular genres (pop, hip-hop, country, etc) to loudness practices at the mixing and mastering stages.

That is why most of the classical CDs and LPs in my collection are over thirty years old.

The reason for the perceived low levels on those early classical recordings is to preserve headroom for the crescendos and other loud movements during tracking without the need for gain-riding(considered a form of dynamic compression in my book).