Getting The Right Loudness Levels and Mix

Following on from Open Mic 10, some questions were raised about optimum loudness levels and how to achieve them. I had been having a side bar conversation with @davidp following his recent OBS experiments and some of this was centred on using loudness meters.

So rather than re-invent the wheel, I’ll let this guy describe how he uses Youlean Loudness Meter 2 which can be obtained as freeware and upgraded to pro at a reasonable price (mho give the tools capabilities). It can be added as a VST plugin to OBS or can run a desktop standalone program. And in your DAW. Youlean Loudness Meter

Another Loudness Meter, which I believe Mr P has been using provides similar functionality, so I will include a link for that as well. Orban Loudness Meter which is completely free but imho less functionality.

I found using the Loudness Meter allowed me to get an good set up much quicker than in the past, with the knowledge on what True Peak level was required and what my target Integrated LUFS level should be for YouTube.

So as an example based on how I have been setting up for the JGC Open Mic recently but could equally be applied to AOVYP uploads, here is what I have been doing.

Audio Input (Guitar/POD Go, Vocals/PlayAcoustic and Drum & Bass BT/Trio+) are connected to my audio interface and fed into OBS as an ASIO Input Capture source. The AIC then has Gain, Limit and Youlean Loudness Meter added (the latter as a VST plugin). With reference to what was said in the above video, I set the limiter to -2 db.

I will then play each individual Audio Input and monitor the Loudness Meter and bring each input up to or near -14 db Integrated LUFS. As the combined input will normally boost the overall and final output volume, I target around -20 to -18 db for the Guitar and backing track inputs. Vocals I aim to get near to -14db. This may require adjusting the AICs Gain to boost each input to avoid clipping on the Audio Interface or any of the FX “boxes”

Once those initially levels are set up I will record a video practice run in OBS with all parts and again monitor the levels and tweak the gain in OBS as required. What is also helpful is to convert the OBS video into an audio format (WAV MP3) and drop it into the Youlean standalone meter on the desktop, as you can see the wave form and levels for the whole song. See screenshot of Should Have Listened recorded in OBS.


So although this shows a couple of sections in excess of -14 db I was happy to leave this song at -16.5 db as I knew my vocals would be louder on the second song.

If you are over or under the target level, its just a case of fine tuning gain to get as close to -14 db as possible.

Interestingly, when the same set up was then recorded in Zoom (from OBS) the overall level dropped. Integrated LUFS for OBS 16.5 and Zoom 19.7. This is something a few of us have suspected for a while but now backed up by evidence. So I could have actually boosted the OBS audio gain by 2-3 db and have still fallen just short of that magical -14 db figure when the audio passed through Zoom. Worth bearing in mind that there is some headroom!

All of the above can be done without using the OBS, you would just use the desktop version of the meter and make gain/level adjustments on your Audio Interface to achieve the same end.

FWIW - in the past many of us have referenced the YouTube Nerd Stat figure of -6 db being the optimum level to aim for (right click on YT screen) and both my OM X songs, recorded using the method above came in at -5.7 db when I did my final Dry Run videos before the show. If I had hit that figure in the past by trial and error, that would be the time to step away and leave everything ready for the show.

Hope folk find this useful. I am in no way a technical expert in these matters and have learnt purely by trial and error and the information shared between community members. If there are any questions I’ll do my best to answer but I am sure more worthy sages than I will offer their valued opinions.

Cheers

Toby
:sunglasses:

9 Likes

Cheers Toby, I would add Audacity to the list, which has an inbuilt level monitor. It’s open source so free.

Available for Windows / Linux /Mac

I’ll check out the videos, I’m never sure if I’m aiming for peak -14dB or average / hovering around -14?

I would say that should be the average for YouTube but other streaming services differ level wise. Spotify (not that I use it) seems to be the same as YT at -14db (at the moment). The True Peak limit should stop any overall clipping and I think YT will dial down volume if its too hot. But as the video says YT does not increase DB like other platforms (Spotify) so getting as close as you can is the target.
:sunglasses:

I haven’t yet put much thought into this aside from just preventing clipping & setting levels between vox & guitar. Preventing clipping at the Focusrite and then preventing clipping in my DAW.

What’s the logic behind the -16, -14, -6 that you mentioned? And how do you factor in dynamics changes, e.g. when you have quiet parts of a song and loud parts?

That’s a little difficult to judge because the meters in Audacity are simple dB meters, which show the instantaneous value.

Most recording apps, like Audacity, and all full DAWs have these as a minimum.

These.are, however, different from LUFS meters. LUFS metering is a standard used to measure perceived loudness, and which involves some signal processing to compute the value. It’s the standard used by YouTube, and most streaming services these days.

There’s no direct and easy translation between simple dB and LUFS metering, so using dB meters will involve a fair bit of guesswork.

It might be best to get one of the meters Toby suggests and use them to give you more useful values.

Otherwise, I believe Audacity has an RMS meter option. Use that and focus on the RMS values, which will be better than using the instantaneous or peak values.

Cheers,

Keith

Cheers Keith will check those out.

JK

This video will make more sense of that where there is reference to the loudness levels, used by Spotify https://artists.spotify.com/en/help/article/loudness-normalization
YouTube would appear to use similar values from what I have seen on various sites.
I was going to include it above but thought it might muddy the waters.
The difference is, Spotify (and others) will raise or lower loudness for consistency but YT will only cap if too loud, not raise. So aim for an Integrated LUFS setting of -14 if uploading to YouTube. I use -16 for wriggle room and also that I can come in a little hot on the OMs because of the style I tend to play or in the last case knowing I was going to let rip on the vox at the end of Turn and I was clipping at the higher level.

Can’t quite remember the source for the optimum YT db level of -6. It was either mentioned here years ago but also in one of Kenny Gioia’s Reaper Mania videos. I’ll see what I can find on that one.

The OM X video is showing -2.9db on the Nerd Stats, so I might create an MP3 version and drop it into Youlean LM and see what the stats are for the whole show. That could give us a better target for the future, assuming @adi_mrok did not tweak the audio levels on the Zoom video before posting ?

Hope that helps a little more.

Cheers

Toby
:sunglasses:

1 Like

Spot on Keith, using the Integrated LUFS metric made things a lot simpler and quicker to achieve a good baseline to start from.

:sunglasses:

1 Like

Not this time around, on all recordings perhaps twice I needed to fiddle with levels.

1 Like

Thanks for confirm Adi.

I dropped an MP3 version of the show into the Youlean Loudness Meter and saw some interesting results. Across the whole show we averaged -17.3 Integrated LUFS, so a little down on the optimum -14 LUFS for YouTube.

But what is interesting is the number of occurrences, when the Max True Peak was exceeded. Ideally this should be below -1db but we average -0.1db. So you peeps are hot !

The screenshot shows the whole show and that True Peak clipping is shown by the red bars at the top of the screen. Each time that happens we are likely to get pops, crackles and distortion, as explained the the videos above. So it may be that’s what we have put down to bandwidth causing audio break up, could just be that we are too hot now and then. I’ll do some DAW comparisons on this and see if the peak clips match with my audio break up on the night.

I found this in relation to the YT Nerd Stats, which kind of half explains what is going on.

Move the volume slide on the YT page and the Volume/Normalized levels will change accordingly. But it still does not explain what a negative or positive db value means ie is the sound short of YTs optimum level or above? Remember YT does not adjust loudness/volume up only down. So I am trying find that answer next.

:sunglasses:

Volume / Normalized

The first part of this stat is simple. The volume shows where the volume slider currently is as a percentage, with 0% being all the way to the left, and 100% being all the way to the right. The second part of this stat is a little trickier to explain, however.

“Normalized” refers to the normalisation of the audio, which is how YouTube protects your ears from drastic volume variation in a video. Think of this as an adjustment. If the normalised volume is 80%, then you turning the volume down to half will make the actual volume 40%, rather than 50%.

The final stat here is the “content loudness” value, which refers to your videos loudness level in comparison to YouTube’s reference level. The reference level is how YouTube ensures that there are no dramatic differences between different videos on their platform, which would be jarring to anyone who is just enjoying a bit of downtime and letting YouTube take the wheel with regards to what they are watching.

It is this value that is used to determine the normalisation amount, as YouTube attempts to bring this video in line with their reference volume. This value will not change based on any action taken by you, such as moving the volume slider. In fact, it is fixed at the point of upload, and would only change if the video was edited to adjust the volume, or YouTube changed their reference volume.

1 Like

Thanks Toby, really comprehensive and useful, appreciate the time you’ve put into posting this!

1 Like

Ok I have managed to find a better explanation here.

full article here

The final value is the “content loudness” value, and indicates the difference between YouTube’s estimate of the loudness and their reference playback level. This value is fixed for each clip, and isn’t affected by the Volume slider.

So for example a reading of 6dB means your video is 6dB louder than YouTube’s reference level, and a 50% normalization adjustment (-6dB) will be applied to compensate. Whereas a negative reading of -3dB, say, means it’s 3 dB lower in level than YouTube’s reference, and no normalization will be applied, so the normalization percentage will always be 100% of the Volume slider’s value - YouTube doesn’t turn up quieter videos.

:sunglasses:

Finally a good video from Kenny Gioia using in YouLean LM in your DAW in his case Reaper but should apply to all DAWs. So this could be useful for folks mixing multi tracks and not just prepping for Open Mics. Mr P may have shared this in the past but relevant to this thread.

I have given up on the chase as to why the Nerd Stat of -6 db Loudness level was recommended in YouTube but it was banded around some years back. But these platforms are changing all the time, so it may well be redundant. Certainly if you hit that figure your YT post or OM audio will have a pretty good sound level but you would have some scope to push a little louder. But if you see a positive number, best to dial it back below zero.

Anyway, plenty of stats, figures and tech info for those who like that stuff but at the end of the day a simple Loudness Meter like the one from Youlean, should mean you can optimise your levels for posting or live performances much quicker - with out the constant need for YT uploads and constant stat checks (which I did in the past).

Hope all this helps.

Toby
:sunglasses:

1 Like

Full nerd mode engaged, going in!

Cheers Toby :+1:

I found that page really good - and much quicker to skim an article for info than watch videos :grinning:. Learned about the “stats for nerds” on youtube videos to work out if it’s been normalised or not.

There’s this page on the same website as well, he gives more advice on how to do loudness levels: How loud ? The simple solution to optimizing playback volume online - and everywhere else - Production Advice.

This bit being the main takeaway for me: "Master no louder than -10 LUFS short-term at the loudest moments (with [True Peaks no higher than -1])"

I’ve downloaded that Youlean plugin and can now check this stuff in future, thanks Toby.

A good article JK. I think if we had been able to put a -1 or -2 db limiter on the shows audio, all those clips on the screenshot would have been avoided. Interestingly, I just drop the OM MP3 into https://www.loudnesspenalty.com/ and the Spotify reading came out bang on the money at 0.0 db. And given YT and Spotify have an ILUFS level of -14 it was spot on for YT.

:sunglasses:

@TheMadman_tobyjenner

Excellent write-up, Toby.

This was news to me. I have used YouLean before and never realised it ran as a standalone. I thought it ran only as a vst in either my DAW or in OBS. That was why I opted for Orban.

I agree with you, YouLean is the better choice, if only for the histogram and ability to review the graph and pick up the moments that exceeded the True Peak threshold.

So I decided to give it a quick test run and run my YT upload through YouLean. That was easy enough.

But the results were surprising. Given the Stats for Nerds result of around -3.5dB I expected Integrated LUFS in the region of -16 to -14. But it was way way softer. Turn on Orban and I see what I expected.

Then I started to fiddle with the volume level on the headphones and that appeared to change the levels in YouLean. That makes no sense to me. Note that I only have the free version.

So some tinkering is going to be required to make sense of that … something for another day.

And I should add, the key for me, is to meter the integrated output of OBS. It is helpful to meter individual sound sources but mostly what I want is a view of the final mix. It is a pity that OBS does not show the master mix anyway and allow one to apply filters to that, like one would in a DAW like Reaper.

Maybe YouLean will work better if it’s audio source is whatever the OBS monitor output is set to be? My expectation would be for Orban and YouLean to be similar.

Or maybe, my brain is still in power-saving mode?

I quite agree, which is effectively what I am doing when using the Xenyx AI for the ASIO Input Capture. Although I can test each “input” individually to get a base setting, when played in realtime, vox/pa, gtr/Pod and Trio/Bass n Drums, its all presented from the Xenyx as a single stereo output. That way I get to see the whole picture. The UMC1820 would give me individual AIC “tracks” in OBS and I would have to guesstimate how the sum of the inputs equates to the single OBS output. The only way to get round that is to then run Youlean in standalone on the desktop, so it picks up the OBS output from the soundcard. ie what is sent to the PC speakers/headphones. That sounds right in theory but yet to test, as I will hit latency issues using the UMC. However, it could be tested with pre-recorded tracks in Reaper and use them as the AIC for OBS. But life is then getting complicated again.

But I could monitor the UMCs dry mix so I am latency free and Youlean run in standalone should get the delayed “wet” PC output. So good enough to monitor maybe ?

Re volume fiddle, was that via the Scarlett ? Seems a bit odd that it would raise the Luffs, so yet another mystery.

After a few weeks off, I need to get back into basic playing before exploring this rabbit hole again. All this can wait and for you too my friend !

Cheers

Toby
:sunglasses:

@DavidP. Put that Klingon dictionary away! You need rest!
@TheMadman_tobyjenner. Don’t encourage him! :smiley:

I was hoping he’d take notice of my last comment !! :rofl: