How to hit target Integrated LUFS?

I need to deliver a 5.1 film mix within these specs:

-7db thru -10db, +/-1db & LUFS @ -24, +/-1

But when I run my Loudness Track Quick Analysis, I get an Integrated LUFS of -27.6 and a range of 18.6 LU

I know I can Normalize to target LUFS in the export dialog box, but my stems should sum to same loudness no? So what’s my best option? Raise the level of my groups a little and run the quick analysis again? I also need to insert tones for color bars and don’t want that level messed with. Tell me what the pros are doing!

Well, I’ve never actually done a feature 5.1 film to specs, only broadcast content (i.e. TV shows / commercials)… but anyway…

I’m not sure what the specs mean. I’d be a bit worried if they’re written exactly like that. I usually want crystal clear specs with a clear hard maximum total peak and an integrated loudness target with the variation range.

Since I guess you’re really asking about the integrated value and not peaks I suppose I’d say that the best thing to do is to mix it until you hit it. The problem with ‘lifting’ it all is that if you have peaks right at the limit then as you’re lifting your mix they’ll either go over or get squashed by a limiter, which means you’ll have to audition the peaks again anyway so you might as well run another mix pass (?).

I have done what you suggest; simply raise the groups on which I have limiters, and then I try to keep an ear and eye out for peaks that are objectionable. Depending on the content and project though I’ve sometimes preferred to use either a VCA (obviously in Pro Tools ;-( or a set of linked faders (groups/auxes), set them to latch trim, and then just run the program top to bottom. As it’s running I gently adjust levels throughout. It works if you’re careful and understand where you can lift levels. I guess the point I’m making is that if you do it this way and can foresee everything then you can hit the target pretty much right on as opposed to having to “guess” what +XdB on all groups really adds up to, because it might not be enough or it might be too much… and it can be smooth and undetectable. Of course I wouldn’t really do this with a sensitive mix that has been approved. That’d be a different animal I think.

Did that make sense?.. I’m not sure…

Thanks for your input. I don’t have any limiters anywhere so I might try raising the groups and see what that does, how it sounds and check the peaks again (only a few gunshots and an earthquake). Would love for it to be a 30 sec spot as a new pass would be a lot quicker!

Normally I think it is requested to deliver a program that has a specific integrated loudness value, and a max true peak value.

Nuendo allows you to export and normalize to such values. When exporting, in the dialog, in the File Format section there is the checkbox, normalize to integrated loudness and mac true peak level. You can set these. At export time, Nuendo is analyzing the complete material and then making it louder or quieter so you get the perfect match.

The loudness module in RX5 does something similar, though I think it’s not only normalizing but also transparent compression. If your material is louder, it compresses it slightly in the loudest areas. Also to match to specification and you’ll get the perfect file.

I’m not sure if you meant that?

That’s what I’d like yes. However, I’d like my stems to sum to the same level, which they would not if the Mix was normalized/compressed and the stems are not. That’s why I thought playing with the level of my groups would work. Thanks for the RX5 tip. I’ll look into it.

Do NOT normalize, unless you are close to the target.
I put the brickwall limiter in each output buss and adjust the Group “pre” values to get closer to the target.
But … you Always have to re-check your mix. (That is if it invloves more than 2 or3 LUFS)
Because it breaks the fundamental “reference” of mixing at a fixed level.
So if you are monitoring @ 76dB, and you raise or lower the complete mix level, some things will be way too loud, or some thing will fall below the hearing thereshold.

Unless you are making an old fashioned “squashed/compressed/limited” mix (which we rarely do in Post), NEVER normalize to your target loudness.
(Unless it’s only a few LUFS)

In your case, raising the level from -27LUFS to -24 LUFS will very probably make that your peaks hit the brickwall limiter. Very probably your dialog will be way too loud. Etc …
Set your reference level correct. Adjust your dialog to a comfortable level. Mix the rest around it.

Fredo

Fredo, are you talking specifically about stems? If you export stems and normalize, then you’re going to get a mix that’s out of balance. But if you have a complete mix and only export that, then normalizing should anyway not be far off.

Aah, and there’s another problem I see now. When you have dialog and a lot of quiet passages in a movie that brings down integrated loudness and you normalize, and the normalization process lifts the loudness by 10dB, then your dialog will be loud. I get it.

I used it for voice-over only tracks for games, where I export lots of single short VO clips for one character and there, it’s mostly helpful unless a character has a passage where he/she whispers. That’s not so good then and needs to be adjusted individually. Or I also use it for sound effects to have a consistent level and then I can adjust the volume in the game engine. Make loud explosions louder or other sounds lower in volume but the normalization gives me a good starting point and reference.

Thanks, I’ll try that approach.

Hello Sunshy,
one thing seems not to be clear: Is it a finished mix consisting of stems that you have to deliver or is it a mix from scratch? If you monitor your mix from the start of your mix session than it is no problem to target to the requiered electric value and if you want to make use of the allowed value you have to correct your mix not by raising or lowering the whole mix but certain parts or elements. If you have an existing mix consisting of different stems rerun the mix by adjusting the single stems and monitor the level of the overall mix. But to achieve a proper result you have to use your ears anyway.
Carsten

Two examples, just for the sake of making my point.
We assume that you a re mixing at a “correct” monitoring level, which is the way it is done in broadcast.

-Suppose you have mixed an explosion at the limit of what your ears can handle.
-Supose you have mixed low level content to the point where you can just slightly notice them without actually clearly hearing them.
-Suppose you have mixed the loudest shout/cry/dialog just below the point of hurting your ears, and the lowest dialig level to where it is just underdtandable;

(Above examples are typical and exemplary of how we mix in broadcast without using metering)

Then try to imagine what will happen to your mix when you raise the level 3 to 4 dB.

Disclaimer; when you mix the old fashioned way (compressing and limiting everything so everything “mixes itself”, then normalization will work perfectly)

Fredo

Keep in mind though that maximum True Peak varies from broadcaster to broadcaster, regardless of national specs. Here in the US I’ve seen everything from -2dBFS to -10dBFS. The latter obviously makes little sense at all these days, but some don’t care.

So anyway, I’m just sort of saying that if you have a brickwall limiter at -10dBFS, and your loudest item is at -10.5 for example, then as you boost your level by those 3 to 4 dB it’ll never be that loud, it’ll just be compressed. Not that that’s not a problem, just a different one.