What's the best way to make sure podcast audio is the right volume?

[Southpaw1496]

I edit a podcast with a lot of different speakers, none of whom are professionals, which means that they all have very different volumes, and also that the volume of each speaker can vary throughout the recording, so making sure the volume is consistent throughout the recording is both an important and daunting task.

From what I've read online and observed by measuring podcasts I listen to, the standard volume for podcasts is -19LUFS. Currently, to achieve that, I first strip the silence from all the tracks (which works very well for me, despite what I've read about it), then select all the regions and use the Normalise Region Gain function to set them all to -19LUFS, and it's here where I've recently started running into walls. When using Normalise Region Gain, you can either base it off the Collective Selection, Individual Tracks, or Individual Regions. Individual Tracks hasn't really worked for me, as the volume of each track is quite variable, so most of it ends up being at around -24 to -22LUFS, aside from a few louder bits that bring it up to -19 eventually. Individual Regions works better, but it has a nasty habit of making very small regions (such as censor bleeps) very loud, and also can make the peak up to +5dB at times, which is not ideal.

I've realised I need a new solution, but I don't know what a better way to do what I need is. I briefly tried to use the compressor, but all the tutorials and explanations I've found are either very brief or related to music instead of spoken word. I also haven't found a tutorial for "There's multiple people, and we want them to all be the same volume". I haven't found a good tutorial on properly using Normalise Region Gain either, and the fact that no-one seems to use it for this makes me wonder if I should be.

 

Does anyone have any advice?

[David Nahmani]

3 hours ago, Southpaw1496 said:

it has a nasty habit of making very small regions (such as censor bleeps) very loud, and also can make the peak up to +5dB at times, which is not ideal.

Move all your censor bleeps to their own track so that you can easily adjust their volume separately from the rest of the speaker tracks.

How is your project laid out? Ideally each speaker should have their own track, and you can first normalize all the regions in the project, then balance the different speakers (if one is still louder than the others then deal with it by turning down his channel strip fader), then balance at the individual region level using either region gain or volume automation.

It's also a good idea to use one or two compressors on each speaker's channel strip for level consistency. 

Here are a couple of related topics that should help you practice your compression chops:

 

[Southpaw1496]

1 minute ago, David Nahmani said:

How is your project laid out?

Each speaker does have their own track, and currently Normalise Region Gain is doing a pretty good job of keeping them at generally around the same volume, so I'm not worried about that. Partly I would like to understand why using the normalisation function does things like increasing the gain with very small regions and going so high over the 0dB peak, instead of just giving up and trying to work around it, because it seems like it would be a valuable tool if only I knew how it worked and could therefore utilise it properly. Are these issues due to something about the way LUFS is calculated, or is it some other reason?

 

Thank you for the resources on compression, I'll have a look.

[David Nahmani]

2 hours ago, Southpaw1496 said:

Each speaker does have their own track, and currently Normalise Region Gain is doing a pretty good job of keeping them at generally around the same volume, so I'm not worried about that.

Ok that's great. 

2 hours ago, Southpaw1496 said:

I would like to understand why using the normalisation function does things like increasing the gain with very small regions and going so high over the 0dB peak

Basically the function adjusts the gain of each region so that the region reaches the set target integrated LUFS value. LUFS is a completely different way to measure a signal than peak though so regions with different material will have completely different behavior as far as the peak/LUFS relationship. Also integrated LUFS being some way to measure the average loudness over a period of time, shorter regions may have a more "volatile" measurement vs longer ones. 

In any case, no tool or measurement will give you a reading of the exact loudness sensation, and while they can do the bulk of the work for you, in the end you'll have to use your ears to fine tune the results. 

[Southpaw1496]

So the best solution is just to use Normalise Region Gain and then go through it to make sure that everything looks OK? What about making sure the peak doesn't exceed 0dB, is that something a compressor can help with?

[David Nahmani]

A limiter is normally the tool used to avoid peaking above 0 dBFS, however you can use a compressor to make sure the limiter doesn't work too hard... basically those are all dynamic tools that lower your signal when that signal reaches a certain threshold. If your signal is real loud and you slap a limiter then the signal won't go above 0 dBFS however the limiter will probably be distorting the signal. So if you have a compressor that says "anything above -5 dBFS gets reduced by a certain amount" for example then the limiter will work a little less. 

Keep in mind that both the compressor and the limiter will affect the LUFS readings, so make sure you meter the loudness on the last slot of the Stereo Out channel strip. 

[Southpaw1496]

Is that this area?

[David Nahmani]

Yes!

[Southpaw1496]

One more question, should I be applying multiple compressors, does it matter which one I use, and should I be applying them per-track or is it OK to apply one for the entire stereo-out?

[David Nahmani]

There's no rule, you apply compression when and where needed. Logic's compressor is a great compressor and for this application I would just use the default "Platinum" model, which doesn't apply any sound coloring. 

I would recommend starting with using one compressor for each track and another one on the stereo out. 

A good starting point to dial in each compressor is to turn off the auto gain, set the ratio to around 4:1 and adjust the threshold so that some of the weaker useful sounds make the needle move ever so slightly. What I mean is that let's say you have a speaker who is saying "I love chocolate" and first you hear him breathing in, then he says "I" rather softly, then over-emphasize "love" which is very loud, then "chocolate" is at a normal volume. Ideally the needle should not move when he's breathing, should barely move when he says "I", move a lot when he says "loud" (lot of volume reduction) and move a little when he says "chocolate". 

Hope that made sense? 

[Southpaw1496]

It did, thank you

[David Nahmani]

Great, you're welcome. 😀

Next, if you use a compressor on the Stereo Out then its role is more for the overall volume of the piece to stay constant, so that if everybody is rather soft at 1mn05 and everybody starts shouting at 1mn23 then the compressor should reduce the level at 1mn23 and be barely moving at 1mn05. Same idea but applied to the global mix of all the speakers rather than for an individual speaker. 

Note that if you've done a great job of normalizing and compressing the individual tracks, you may not need that compressor on the Stereo Out, or it may be working very little.