Levelling out track amplification for a 1 hour Podcast.

Hi there,

I have read multiple posts and watched many YouTube tutorials but I can’t seem to identify the best way to level out a long audio Podcast track with two people speaking without manually going and looking for the short random peak spikes where it’s typically a Phhh or an SSSSS spiking in sections of the track. When you normalise a 1 hour track it brings everything down judged on the track where these highest peaks are. After I normalise I then go through a compression process but the lows of the audio are still too low. I believe it’s the random Phhh and SSSSSS peaks I have throughout the track that are causing the issue. What is the best way to level out a long track in an automated fashion so the audio is nice and level? Some of the video tutorials I’ve seen use a “clap” as an example but then they manually identity these clap spikes and reduce them before going through the normalisation process. Ideally I’d like Audacity to automate that process as my spikes aren’t all in one place like the clapping example.

Many thanks,
Cameron

Stop doing one hour recordings.

It’s hard because you need to do some manual selection and filtering. In a 10 min. file, you can easily do that.

A one hour recording hardly moves.

Every source is different. There is no best way. I was just trying some new denoising gizmo’s and there is no “universal” solution. One works very well for most sources, as long as the noise is not too bad and repetitive. Automagical, even. The other isn’t useful at all, unless you want to spend some days with it.

In general, you remove unwanted peaks before normalizing. Otherwise, as you’ve noticed, the normalizer does what it’s supposed to do.

To bring up low level passages, you need an expander, not a compressor. It may sound illogical, but that’s what these are meant to do. You can use some compressors as expanders, but most aren’t easy.

And once you’ve found a procedure that works for your sources, you can make chains. The work is done with the first fragment, the others can be treated by the chain.

I’m typically cleaning up 2-3 hour recordings and the process I have been using looks a little like this:

  • Noise reduction - I switch the track to DB view to do this because it shows the noise better.
  • Limiter - depending on the starting volume this can be down to -10 or not quite that far, basically the goal is to kill the massive peaks.
  • Amplify - This will make everything louder again, but now your peaks are much smaller it will make the quiet parts louder.

However just this week I’ve had the RMS-Normalise plug-in recommended which I’m putting in this process instead of Amplify and seems to work pretty well as an alternative.

De-essing selectively reduces those.
There is an Audacity-specific De-esser plugin, but it’s comparatively difficult to use,
( you may be able to find a more user-friendly de-esser plugin elsewhere)

Do you have access to the individual voices? If you do, you can use the AudioBook Mastering process. That’s where RMS Normalize originated.

https://forum.audacityteam.org/t/audiobook-mastering-version-4/45908/1

The job of the mastering suite is to level out volume variations and put the presentation volume exactly where the publisher wants it. It uses a trick you may not be used to. Sound doesn’t overload inside Audacity. No matter how loud the sound gets or what happens to the blue waves, you can bring the volume back down with no damage (don’t try this outside of Audacity).

RMS Normalize depends on that. RMS Normalize will put the voice volume exactly where it needs to be and doesn’t pay attention to the blue peaks. Effect > Limiter comes along and squashes the peaks so they “fit.” The only other variation is noise and that is a paper all is own.

The tools in the mastering suite depend on each other, so for at least the first trip, follow the recipe exactly.

You may find that if you stop using Noise Reduction, the aggressive SSSs and sibilence isn’t as bad. Making a noisy recording is to be avoided at the recording stage. Fixing noise in post-production is not pleasant or fun.

Koz

I’d cut these in bite sized pieces, if possible. Maybe that’s just the way I’m used to doing stuff?

and the process I have been using looks a little like this:

  • Noise reduction - I switch the track to DB view to do this because it shows the noise better.
  • Limiter - depending on the starting volume this can be down to -10 or not quite that far, basically the goal is to kill the massive peaks.
  • Amplify - This will make everything louder again, but now your peaks are much smaller it will make the quiet parts louder.

When you have compressed, you shouldn’t have to use an extra amplify step. But that’s just nitpicking, as gain doesn’t “do” anything to the sound.

A good read, if you have some time to spare, is the stuff around EBU metering. A lot of people swear by 16 dB LUFS for podcasts.

FWIW I always work with the dB scale. % seems unnatural.

However just this week I’ve had the > RMS-Normalise > plug-in recommended which I’m putting in this process instead of Amplify and seems to work pretty well as an alternative.

That works fine. It was created for voice-over, I believe, but I suppose it matches a lot of podcasting work too.

Thanks @cyrano - I clean up the whole recording because that way I know I have a consistent sound and it’s a pretty quick job for me, although my PC has to think about it for a while. If I break it up there’s a pretty good chance that I will end up with different sections sounding different because I have missed a step or done something differently. Once it is cleaned up I split it out into hour(ish) sections for editing down into episodes.

So if my basic recording level is somewhere around -10 to -12 db for average speech and I have limited it down to -10 to reduce the peaks are you saying that amplifying doesn’t make it louder? I don’t understand how that works - it sounds louder to me? Also the peaks are higher, which I took to imply relative loudness. The goal for me is for the five ( or more ) tracks everyone has recorded locally on their own machines to be as close to equivalent in loudness as possible. They can certainly look the same on the waves without being the same to the ear.

When I was using direct compression rather than limiter-amplification it just seemed to make everyone sound super-robotic if I had a strong compression or make no difference at all with a lower level one, which is why I have moved towards limiting and boosting instead. The RMS normalise makes sense seeing as I want vocal clarity from a set of people speaking, which is basically a pretty similar goal to recording an audiobook, except busier…

There’s a relatively recent voice leveler. I need to go looking for it.

My go-to generic leveler is Chris Compressor.

https://theaudacitytopodcast.com/chriss-dynamic-compressor-plugin-for-audacity/

I change the first value Compression Ratio from the default 0.5 to 0.77 and it’s a dead ringer for the local FM station. I process the podcast version of a two-person radio show. On the radio, their performance goes through the transmitter compressors and it sounds like every other radio show, but in real life, it’s almost unlistenable from volume changes.

Looking.

Koz

Here is the post with AGC (Automatic Gain Control).

https://forum.audacityteam.org/t/agc-automatic-gain-control/26706/37

That’s still not the one I was looking for, but that may do it for you.

Koz