Let's put an end to this loudness war now, shall we?

Hi everyone!

Sorry in advance for the long and technical post but this is really important if the AJ ecosystem wishes to maintain its competitive edge in the light of (not so) recent developments.

As we all know, it is quite common among AJ authors to maximize the volume of their tracks with compression/limiting. The logic behind this is that, other things being equal, we humans tend to gravitate towards louder stuff. When a customer is browsing through search results, if your track is louder than mine, you have a higher chance of selling your item.

This phenomenon has been a major issue in the music industry for a long time, and itā€™s called ā€œloudness warā€. What makes it an issue is that when you squash your music to make it louder than everyone elseā€™s, you compromise its dynamic range and it usually ends up sounding harsh as well.

Pros in the music industry have been preaching against loudness war for years, and their efforts are finally paying off. Many leading digital platforms have in recent years started to implement their own normalisation algorithms in an effort to maintain consistent volume from one song (or video) to the next.

Youtube, for example, normalizes everything to around -13 LUFS, which is very quiet compared to what most authors offer here. What does this mean? Iā€™m glad you asked.

Take a look at this screenshot from a search result I grabbed a couple of minutes ago:

forum_lufs
The track in the middle is mine, and it conforms to Youtubeā€™s new standard. Meaning, when someone purchases it and uses it in a video, Youtube will not raise or lower its volume. It will remain the same, and sound the same.

The track above it, however, has a loudness level of ~ -8 LUFS. Youtube will turn it down significantly, and it will end up sounding much weaker. Imagine what that waveform would look like if you cut it in half horizontally. Thatā€™s what it will sound like as well.

If youā€™d like to know more about how and why this is important, and how you can measure LUFS values for your own tracks, there are excellent articles available all over the internet. Thereā€™s an excellent one here.

Long story short, as we try to out-maximize eachotherā€™s tracks for attention here on AJ, we are inevitably violating new standards and causing potential confusion and disappointment for customers. If AJā€™s competitors take the lead in embracing these standards, they will definitely have an edge. No one wants to upload a video and then find out that the music sounds almost half as quiet as it did on their video editing software.

So what is the solution? The best solution would perhaps be for Envato to implement some sort of automated normalization on uploaded tracks, or to impose a standard on authors. The next best thing would be for us authors to reach a gentlemenā€™s agreement to end this war and to observe these standards for the benefit of everyone. It is also important to educate customers that the loudness war is (mostly) over. Iā€™m planning to mention this in my item descriptions in the future, but with my tiny little portfolio, that wonā€™t really mean much :slight_smile:

15 Likes

This is interesting, I had no idea about this and I often buy audiojungle music. Thanks for the info!

:hugs:

1 Like

Thank you very much for the invaluable information :slight_smile:

I did not even know that Logic Pro X had a loudness meter plugin.

1 Like

This issue has been mentioned a number of times. I think the problem is to go through 700k+ approved items, which have up to 5 versions .wav and .mp3 plus preview file.
Also usually our music is in background, so heavily compressed track is not twice as quiet as not a compressed one.
That being said, Iā€™m all for introducing LUFS restriction, itā€™s just difficult to do.
Cheers!

While thereā€™s no denying the top example is extremely loud, itā€™s important to note that it contains 5 separate files compared to individual files for the other examples, so it appears much more compressed than it would if it were only one file.

1 Like

If an automated system is implemented, it should work on previously added tracks as well. Youtube probably had gazillions of videos by the time they implemented normalization, but most of their back catalogue has been processed already. Right click on a video, go to ā€œStats for nerdsā€ and the ā€œVolume/Normalizedā€ line will tell you how much normalization is applied, if any.

Being a nerd :slight_smile: I actually downloaded the preview for that track and measured it before I wrote it has a loudness level of -8 LUFS.

I am with you hereā€¦
I recently decided to lower my loudness on my own beat selling web site, and my goal is around -12. Still louder than -14 but wellā€¦

Now on Audiojungle, I see that most of my direct competition is at -9LUFS or lessā€¦ so I decided to stick with -10dB which is a value I get without having to squash things (in the hiphop genre at least).

But then yes, when uploading my track to youtube, Iā€™ll have a problemā€¦ this mean I would have to export several versions for each of my tracksā€¦ a nightmareā€¦

Now, I had the idea that we might have a preview that is a bit louder, and that in the zip you can find less loud musicā€¦

But the ideal would be that yes, audiojungle player applies a correction similar to youtube.
I should calculate the integrated LUFS and lower the volume if needed.

It would be great if the preview playback volumes were all normalised on AudioJungle, as this would remove the incentive to over-compress tracks with the aim of making them stand out against the competition when potential buyers are playing tracks alongside each other.

However, I do wonder if this will be seen as being worth the time and trouble involved. As people have pointed out, our music generally plays in the background and in these cases everything overall will probably sound roughly the same volume as the sound in other videos (which is the point of loudness normalisation). I donā€™t know how much difference improving dynamic range of the background music would really make.

That being said, if the music is the main feature it will not sound as punchy and engaging on YouTube if it has been over compressed.

I thought about this too, but 1) I donā€™t know if this is allowed on AJ and 2) Most video producers Iā€™ve worked with download the preview, do all their editing with it, and once the client approves the music they license the track and just replace it. If they worked on volume levels and automation for voiceover, sfx, etc. with the preview, a different loudness level on the main track could mess things up. It means extra work for them. Or they might not notice it and upload it just like that, with music quieter than they intended (we all know about people who upload videos with watermarked music although they have legit licenses, right?). Itā€™s risky either way.

Consider the typical corporate track, with dotted eighth delays, four on the floor kicks and everything. Letā€™s say this track will be used in a corporate presentation. We see the company headquarters fancy exterior for 5 seconds, during which the music track is heard all by itself, and as we go indoors a voiceover starts to talk about the company.

Now typically a video editor will lay down the music track, and when the voiceover comes in at 5" (s)he will turn the music down (either by volume automation or ducking). In this case the original loudness of the music track will effectively become the reference by which all other audio channels are mixed. If the music track is too loud, the whole mix will be too loud, and will end up being turned down by Youtube.

Not to mention typography videos with stomp/clap music or trailers with epic music where music is the only audio.

1 Like

Such system would only normalise preview files. Imagine a client working with a preview, then changing it to non watermarked version and getting his speakers blown. I doubt there is a way to normalise files in ZIP archive.

I apologise if itā€™s just me being a little slow today, but Iā€™m struggling to see what the problem is here. Youtube turns the audio down relative to the authorā€™s original wavs, sure, but the customerā€™s end product on youtube will be normalised to be the exact same loudness as every other video on youtube. So why does it matter how loud the original files are?

Edit: Furthermore, donā€™t video editors (professional ones at least) normalise the audio in their videos to some kind of standard anyway? (Genuine question, Iā€™m just speculating)

ā€œSuch system would only normalise preview files. Imagine a client working with a preview, then changing it to non watermarked version and getting his speakers blownā€

It would only normalise the playback volume on AudioJungle. The downloaded preview files would still be exactly the same as before.

1 Like

ā€œā€¦ the customerā€™s end product on youtube will be normalised to be the exact same loudness as every other video on youtube. So why does it matter how loud the original files are?ā€

I think you are basically right. Everything will end up sounding the same volume on YouTube regardless of how compressed the original sound is. If thereā€™s an issue, itā€™s the fact that heavily compressed music that is turned down can sound rather flat compared with less compressed music that hasnā€™t been turned down so much.

2 Likes

Thatā€™s right. I pasted a link to an article in the original post, where they explain this visually, using waveforms. A heavily compressed/maximized track truly suffers when it is normalized.

Hi. I am nerd too :slight_smile:

I used https://github.com/danrossi/loudness-cli to get the gain value for LUFS -8 from some 53 of top selling tunes on AJ which were ā€œrecentlyā€ picked ā€œrandomlyā€. Average gain was approx. 1.53. [edit] sorry it was approx. 1.178. I am not familiar with Apple Numbers.app

A table chart of gain values of respective samples can be found at: https://pastebin.com/4uYCKYJ8

Blowing their speakers would kill the deal indeed :slight_smile: But is it too hard to write a script that would unzip -> normalize -> zip again?

Iā€™m not sure I understand. So you set -8 LUFS as the target and the average gain turned out to be 1.178?

Yes. I am not quite familiar with the subject. The script, loudness-cli.py, returns values like:

LRA: 4.3
LRA High: -11.3
I: -12.6
LRA Low: -15.6
LRA Threshold: -32.7
gain: 1.6982436524617444
I Threshold: -22.7

The average value is the average of the 53 tracks on AJ.

Is using gain as a reference totally off-base?