de-Amplify and Adjustable Fade in Audacity - quieting 10 loud unclipped seconds?

[chiefscribe]

How would you handle postprocessing in this situation?

A show I recorded recently includes a brief segment where the artist invited the audience to scream for ten seconds, and they did. 

Fortunately my levels were set so that there's no clipping, but that ten seconds of screaming is significantly louder than the rest of the show.

I don't normalize, but I sometimes use Amplify in Audacity to set the maximum dB to -0.06. 

If the loud part were music I'd leave it as is and just live with the rest of the recording being relatively quiet, but since it's just audience screaming I am wondering about using Amplify to make that part peak at the same level as the loudest part of the rest of the show (-8.09 dB would do the trick from my experiments). 

When I duplicated the project to experiment with this, the onset of the screaming sounds fine because there's a quiet pause beforehand, but there's a noticeable shift when the de-Amplified part ends - the background noise suddenly gets louder.  I was trying to use Adjustable Fade in Audacity to smooth this over but I don't have any experience using that effect and don't know the best practices for how much to select and whether to select a longer span of time then progressively shorter periods and S-fade in or just do it once, and how to figure out the percentage to start with. 

The little shift is not during any music or talking, so it's not a terrible tragedy if it has to stay as is, but I would love to make it smoother if possible. 

Any advice for a newbie? 

[chiefscribe]

or should I be using the Limiter?

[admkrk]

I would use a volume envelope.

https://manual.audacityteam.org/man/envelope_tool.html

[AbbyTaper]

Or...amplify the parts before and after the screaming section so that they peak at the same level as the scream.   Or just cut that section out entirely.

[nassau73]

I would normally handle something like this in RX and I don't use Audacity.

However, you might try something RX like. If you look at the Spectrogram view for that section, can you identify the audience screams? If so, you can try highlighting/selecting the screams and attenuate/lower the volume on the selection.

After that, if there is band chatter or something, creating and increasing the envelope as someone else mentioned would help. I think Audacity works with envelopes the same way as RX. So you can create several "points" both before and after the segment and slowly raise the envelope at one end and decrease on the other to make things sound smoother. I've had shows where the between song banter was much lower than the music/vocal selections but it was important to the flow of the show. It took a bunch of time to handle each break but worth it in the end.

[chiefscribe]

>>amplify the parts before and after the screaming section so that they peak at the same level as the scream

The sound is the entire audience being invited to do a very loud scream for ten seconds, not individual screams that can be isolated. 

[admkrk]

Did you try using an envelope?

I do not use Audacity, so I cannot say anything about how to do it in that application. It is a simple process regardless. Set a pair of points where you want the fade in and fade out to start and stop. Lower the loud section so it is about equal to the rest of the audio. You should be able to eyeball it pretty close. Give it a listen and adjust the points until it has a good flow. Once it is all equal, adjust the overall gain as you like.

AbbyTaper's way would work just the same, but be more work. Rx is overkill for something this simple. Do not over think it.

[AbbyTaper]

>>amplify the parts before and after the screaming section so that they peak at the same level as the scream

The sound is the entire audience being invited to do a very loud scream for ten seconds, not individual screams that can be isolated.

HUH?  You have a 10 second section of screaming, which should be obvious on the WAV form.  Highlight the section before the screaming and amplify it, then highlight the section after the screaming and isolate it. 

If you want, send me the file and I'll fix it for you.

[chiefscribe]

As mentioned in the initial post, the transition after the screaming is harder to isolate without a noticeable abrupt increase in background volume. There is a pause before the screaming, so that end of it is easy to isolate, but talking and music starts while the screams are still happening. I have tried various fading techniques but it still sounds like all the sudden someone turned up the volume on the background noise. 

[AbbyTaper]

Well...you have nothing to lose by sending it to me.  If I fix it, great.  If I can't, nothing lost except my time.

[morst]

HUH?  You have a 10 second section of screaming, which should be obvious on the WAV form.  Highlight the section before the screaming and amplify it, then highlight the section after the screaming and isolate it. 

This would work if you crossfade, but the envelope tool is perfect for this, as admkrk says.

[chiefscribe]

Thanks, will try to get up to speed on using the envelope tool.

The best way I can describe the abrupt bump up in background noise is that it sounds like you are wearing noise-cancelling headphones and they suddenly turn off and you hear the sea of ambient sound appear out of nowhere like you walked outside or through a door from a quiet room to a room full of people.

[morst]

Thanks, will try to get up to speed on using the envelope tool.

In audacity, F5 turns the envelope tool on, F1 goes back to the regular "I-Beam" cursor you might be more used to.


I suggest starting at 1x vertical zoom, and adding a couple envelope points at full volume, one right before and one right after the section you wish to change.
Zoom the track (mouse to the bottom edge and use the double-headed vertical arrow tool to increase track size) so it takes up as much of the screen as possible, this way you will get the best vertical resolution while using the tool. If you are working on a stereo track, you only have to add the points for left or right so you can zoom one channel larger, as in my example image.
When you add the points to make the envelope do its thing and lower the level of the loud section, you'll probably want to give that section the same loudness throughout. The vertical resolution we gained by zooming the track enables you to be sure that the modified section is as horizontal as possible.


I hope this is clear!?



Edit:
Note that the Blue Arrow points to the rectangle with the number values between -2 and +2, and you must RIGHT CLICK in there to get the hidden menu that my illustration shows #HiddenMenu #NotMacLike

[chiefscribe]

Thanks so much for the thorough instructions - will attempt to follow them. Much appreciated.

[admkrk]

That looks like a good explanation morst. There are a ton of youtube videos on this as well. Being able to see the change in the waveform makes it beyond simple. I use Reaper, which does not do that, or not as clearly if you enable it. Peak meters tell you how far off you are anyway.

You do not have to make an hourglass with the envelope though. You can add more points in between to change the fade in or fade out rate. The last set in the attached image has quite a lot of points, for example.