Loudness and normalization process for youtube

Hello, I am learning how to mix audio for youtube videos. I watched probably every video on youtube that deals with normalization and loudness and it seems that everybody has a different approach to this. I hope someone can help me with this.

In my case I have 2 dialog tracks 1 music track and 1 effects track.
Each dialog track is recorded on a different mic by a different person .

What some people suggest is this:

1) First go to project setting and change target loudness to -14LUFS
Then do you audio edit including EQ,compressor including RELATIVE loudness to other tracks.
Once all your edit is done then you can adjust LOUDNESS by trying to get to zero on the meter which is your -14LUFS target and you are also supposed to change metereing transport to BS.1770-4
Then you start adjusting your main to get to somewhere between -3 and 0 on the loudness meter.

OTHER PEOPLE are suggesting this:


Import your clips and NORMALIZE AUDIO LEVELS of your track by changing Normalization mode to 1770-4 and changing target level to -3DBFS. Edit your audio and then at the end adjust main bus to hit zero on the loudness meter as a youtube standard.

So what is the correct approach ?
Do you always set project setting to -14LUFS?
Do you have to normalize clips and don't worry about project setting?
Normalization audio level - Target level is in dBTP and Target loudness is in LKFS not LUFS

It is confusing as hell for me.

Thank you for your help.

Yeah not sure but if you look at 'stats for nerds' YouTube will tell you if it's clamped your audio, and by how much. You can test this for yourself if interested. Ideally I guess, YT would not reduce your audio at all and you alter levels until you get the desired effect

Have you looked at this video?

I do the following,


#1 Cut all clips to length. Normalizing before means you could be getting results skewed by audio you will never use.

# 2 normalize 'background audio' to -24 lufs. Waves, b-roll type stuff. Channel 2

#3 Normalize talking head to -16 lufs, Channel 1

#4 normalize music to -14 lufs. Channel 3

#5 add compressor, EQ, etc to Channel 2 or 3 if needed.

#6 add -1db limiter to Main

#7 do any ducking of music to speech

#9 Balance the three channels to sound nice together

#10 play through whole track and watch loudness, adjust main to get -14lufs target. You could also bounce main to a new track and have resolve analyze loudness, much faster. Delete the bounced track, rinse and repeat until you get -14.


Sucks resolve won't let you normalize the 'main bus' like you can individual tracks. That would make my step #10 take 30s.

Thank you for your help
Zen , could I ask you , does it matter then what your project setting LUF is?

And my other question is , what Normalization mode do you use and if you change Target level .

Thank you .

Hi Charles, I watched that video too, do you think that it shows the correct approach ?
He says not to use normalization because it is an old fashion method.

I do the following,
Sucks resolve won't let you normalize the 'main bus' like you can individual tracks. That would make my step #10 take 30s.

a little tip you might find helpful; you can normalize the main bus by creating a new timeline and drag the other timeline in it, so you basically get one audio/video clip which you can normalize. Or you can go to timeline -> bounce mix to track and normalize the bounced audio.

He says not to use normalization because it is an old fashion method.

Where did he say that?

I only recall him saying that normalizing peaks wasn't of use. And it isn't.

The one thing I never use is normalization. The first thing I do is use clip gain to bring all the clips to a similar working level. Then some eq and compression on dialogue and any fx that need it. As the music has already been mixed I normally leave it alone. I usually use a limiter on the Master bus. It then gets mixed using fader automation with actual faders. Experience tells me where the levels should be, but the final mix level can always be tweaked afterwards before rendering.
Rather than Fairlight I tracklay and mix in Pro Tools but the workflow is similar to Fairlight. The main difference is that I have a calibrated monitoring system so can mix by ear and not watch the meters. The final mix is then brought into Resolve, and the final render done.
Don't get hung up on getting the overall level exact for YT. Nobody is going to reject your video, unlike broadcast where you have to be accurate and have levels exactly right for it to pass quality control.

Yeah not sure but if you look at 'stats for nerds' YouTube will tell you if it's clamped your audio...

YouTube limits audio volume? Gee, you'd never know it from some of the clips (especially the ads...).

He says not to use normalization because it is an old fashion method.

Where did he say that?

I only recall him saying that normalizing peaks wasn't of use. And it isn't.

I thought he was talking about not using normalization at all. But I think I miss understood him.

A very good final test to keep in mind is to upload your video to YouTube (you can leave it as "unlisted" for the sake of testing their compression), and then while watching the video, right click and select "stats for nerds." One of the items listed will tell you how much the volume was changed by YouTube when they processed it. For example, you can see here that the video Charles linked earlier was not reduced in volume when uploaded (100/100).

EDIT: I originally wrote that last sentence incorrectly. I just fixed it.

Thank you for your help
Zen , could I ask you , does it matter then what your project setting LUF is?

And my other question is , what Normalization mode do you use and if you change Target level .

Thank you .

This setting is basically what Fusion uses as 'zero' in the target meter. If you want -16 to be green, set project to -16.

You still have to manually type in the value you want to normalize to.

[

a little tip you might find helpful; you can normalize the main bus by creating a new timeline and drag the other timeline in it, so you basically get one audio/video clip which you can normalize. Or you can go to timeline -> bounce mix to track and normalize the bounced audio.

Yes, kind if a cludge method to me. And you have to 'rebounce' after every edit.

[

a little tip you might find helpful; you can normalize the main bus by creating a new timeline and drag the other timeline in it, so you basically get one audio/video clip which you can normalize. Or you can go to timeline -> bounce mix to track and normalize the bounced audio.

Yes, kind if a cludge method to me. And you have to 'rebounce' after every edit.

It's worth remembering though that the traditional method of mixing was to do it in a room that was calibrated one way or another and where metering would indicate what level you were hitting. So it was the mixing process that got you to the target level. Not retroactive adjustments.

So within that context while it may seem like "a cludge method" what we're really looking at is making it much easier to hit target spec 'retroactively' which allows people less skilled at mixing to do the job. Those of use who mix for a living can typically hit spec by ear in a good room with decent metering. So all in all it's not that it couldn't be easier and less cludgy, it's just that it's far easier today than 10 years ago.