LUFS gate and playback volume

Have an urge to learn, or a calling to teach? Want to share some useful Youtube videos? Do it here!
valankar
Posts: 59
Joined: 24 Sep 2016

Post 11 Jan 2022

This is not really a tutorial or technique, but just something interesting I found recently. Of course, this is not recommended for anyone to do :).

I read this recently:

https://youlean.co/how-to-hack-lufs-normalization/

And was curious if adding a long quiet segment to a song would cause YouTube to play it louder than one without it. As long as the quiet segment is above the LUFS gate described in the article, it does work.

Compare this loud part of a song:

https://www.youtube.com/watch?v=PLFG9QQTWTk&t=203s

To the same mix with a long noise segment at the end:

https://www.youtube.com/watch?v=yUJJDllM43M&t=203s

The stats show the 2nd version is not being lowered in playback volume, but it's obviously louder:

Original:
unmodified.jpg
Modified:
modified.jpg

Again, I don't recommend anyone do this and just wanted to understand how the normalization worked.

What I found works best is volume automation on the final mix. For example, making the non-chorus parts 2 or 3 dB lower than the chorus parts. This effectively lowers the integrated LUFS and results in louder playback compared to without the dynamics. It also sounds better to me when I have those dynamics.
You do not have the required permissions to view the files attached to this post.
Last edited by valankar on 11 Jan 2022, edited 2 times in total.

PhillipOrdonez
Posts: 3084
Joined: 20 Oct 2017
Location: Norway

Post 11 Jan 2022

I don't think it's a good idea to let a normalisation algoritm apply any limiting to your song where you have no control over the results.

In other words: getting turned down is fine. Getting turned up is bad because it will apply limiting to your song and the results may sound worse than you intended.
The latest release:

User avatar
Aosta
Posts: 715
Joined: 26 Jun 2017

Post 11 Jan 2022

If you have a look at my post 'Mastering in the Future' a few posts down there is a video that covers LUFTs, streaming services and loudness :thumbup:


Tend the flame

User avatar
orthodox
RE Developer
Posts: 2278
Joined: 22 Jan 2015
Location: 55°09'24.5"N 37°27'41.4"E

Post 11 Jan 2022

I just downloaded your two clips from YT (m4a format @44.1k), and compared them in Reason with my own LUFS meter.

The short clip:
Integrated = -9.9 LUFS
Range = 3.9 LU

The long clip (only until 4:02)
Integrated = -9.9 LUFS
Range = 3.9 LU

The entire long clip
Integrated = -14.5 LUFS
Range = 18.4 LU

Everything is fine, just as it should be.

User avatar
orthodox
RE Developer
Posts: 2278
Joined: 22 Jan 2015
Location: 55°09'24.5"N 37°27'41.4"E

Post 11 Jan 2022

The histogram shows the part that is below the relative gate and not counted in Integrated loudness in gray color (the whole long clip):

Image

The gray triangle at the top is the position of integrated LUFS.

valankar
Posts: 59
Joined: 24 Sep 2016

Post 22 Jan 2022

Hi all,

I decided to delve a bit more into this to try to fool YouTube into not lowering volume. If you're really bored, join me in this exploration:


User avatar
orthodox
RE Developer
Posts: 2278
Joined: 22 Jan 2015
Location: 55°09'24.5"N 37°27'41.4"E

Post 22 Jan 2022

valankar wrote:
22 Jan 2022
If you're really bored, join me in this exploration:
For more or less uni-level content, a method could be developed.

Suppose we have a 180sec track with Integrated Loudness of -8 LUFS. We would like to append a segment of noise to it that it become -14 LUFS IL. The appended segment (momentary) loudness then could be (no less than) -24 LUFS (-14 - 10, where -10 is the relative gate level by the standard). The length of the appended segment should be:

T2 = T1 * (10^(dI/10) - 1) / 0.9 = 180 * (10^0.6 - 1) / 0.9 = 596 sec

Symbols:
dI = 6 LU, required difference between the original and target integrated loudness
T1 = 180 sec, original track length
T2, length of the appended segment

valankar
Posts: 59
Joined: 24 Sep 2016

Post 23 Jan 2022

orthodox wrote:
22 Jan 2022

For more or less uni-level content, a method could be developed.

Suppose we have a 180sec track with Integrated Loudness of -8 LUFS. We would like to append a segment of noise to it that it become -14 LUFS IL. The appended segment (momentary) loudness then could be (no less than) -24 LUFS (-14 - 10, where -10 is the relative gate level by the standard). The length of the appended segment should be:

T2 = T1 * (10^(dI/10) - 1) / 0.9 = 180 * (10^0.6 - 1) / 0.9 = 596 sec

Symbols:
dI = 6 LU, required difference between the original and target integrated loudness
T1 = 180 sec, original track length
T2, length of the appended segment
Wow, very nice!

User avatar
orthodox
RE Developer
Posts: 2278
Joined: 22 Jan 2015
Location: 55°09'24.5"N 37°27'41.4"E

Post 23 Jan 2022

There is another way. You append a long segment of noise with -24LUFS constant momentary loudness and start playing it all from the beginning until the meter shows -14 LUFS integrated loudness. Then cut off the rest of the noise.

EDIT: You can make a loop in the noise segment and just time it when the meter shows the desired integrated loudness.

valankar
Posts: 59
Joined: 24 Sep 2016

Post 23 Jan 2022

I just found out that Dan Worrall talks about this in his video about "I Won The Loudness War" at around 4:40:



He threatens to do it, but holds back. I tried it with his song and was able to get YouTube to play it back without lowering the volume using above methods. The volume is a bit insane, and I won't link it here for fear of any speaker/ear damage. Message me privately if you want a listen.

User avatar
orthodox
RE Developer
Posts: 2278
Joined: 22 Jan 2015
Location: 55°09'24.5"N 37°27'41.4"E

Post 23 Jan 2022

valankar wrote:
23 Jan 2022
I just found out that Dan Worrall talks about this in his video about "I Won The Loudness War" at around 4:40:
He may have won a fight, but not the war. To prevent total bacchanal, YT could just introduce a limit on max short-term loudness.

valankar
Posts: 59
Joined: 24 Sep 2016

Post 24 Jan 2022

orthodox wrote:
23 Jan 2022
He may have won a fight, but not the war. To prevent total bacchanal, YT could just introduce a limit on max short-term loudness.
The original song is around +2 LUFS:



I did a version where I lowered it to -5.5 LUFS and had YouTube play it without lowering it:

httpz://youtu.be/uFN8tTcEzgI

Then I did a version where I kept it at +2 LUFS:

httpz://youtu.be/wnFbgrpMnXU

(fix the url to actually go to them, and lower your volume)

It seems like you can have original audio over 0 dB with 32bit FP, but perhaps this just all gets to distortion at that point after YT compression?

EDIT: Ah, YouTube came in and lowered the last one, just took some time. Seems like it is doing some 2nd-pass audio lowering that is not related to integrated LUFS.

User avatar
orthodox
RE Developer
Posts: 2278
Joined: 22 Jan 2015
Location: 55°09'24.5"N 37°27'41.4"E

Post 24 Jan 2022

valankar wrote:
24 Jan 2022
It seems like you can have original audio over 0 dB with 32bit FP, but perhaps this just all gets to distortion at that point after YT compression?
Even in the past Loudness War there were rules. You might increase the loudness or RMS by compressing the audio as much as you could, but never, NEVER ever you could peak over 0 dBFS. This rule is still in effect, you can't expect the correct playback of an audio with True Peak over 0 dBFS. I don't see any point in uploading such content, it will certainly clip on a player, if not earlier in conversion algorithms of the streaming platform.
You can still raise the loudness, but there is a technical limit to it, because the peak is limited.

  • Information
  • Who is online

    Users browsing this forum: CommonCrawl [Bot] and 1 guest