Topic: Best way to remove/reduce between-song "audience interference" in Audition?

[jj69]

This has probably been asked before, but I'm not sure what to search for in the archives. 

I recorded a fusion trio in a great room.  The recording is amazing, but for the nimrod seated directly to my left who had what I like to refer to as "loudest guy in the room" syndrome.  He was just one of those guys who has to clap louder and cheer louder than everyone else.  So following each song (and following each solo), I have this guy screaming, "WHOOO!" at the top of his lungs, and clapping like thunder.  The volume discrepancy between this guy and the music is staggering.  In some places he's probably 14db louder than the loudest peak.  The recording can't even be effectively normalized as a result. 

Removal of the clapping is easy enough using Audition's Click/Pop remover.  The screaming, however, is another story.

What is the best tool to use to at least reduce the volume of this guy's voice?  In the past, I've used the "Hard Limit" command with mixed results.  What is the preferred method?  Some kind of compression? 

For the record, I'm using Audition 3.0.  I also have iZotope RX2. 

[vanark]

I have good results using an envelope tool in Wavelab.  It is somewhat manual, but very effective.

[tim in jersey]

I have good results using an envelope tool in Wavelab.  It is somewhat manual, but very effective.

Ditto, although I use Audacity.

Some of the more "respectable" acts I record, like MMW and Bela Fleck and the Flecktones, are in reserved seating theaters where the crowd is very respectful and whisper quiet while the band is playing even during the most quiet portions of their music. But the audience reaction/applause between songs is thunderous even compared to the loudest portions of the music itself.

I apply the envelope tool to the applause prior to normalizing/amplifying the entire wav, and in some cases I've been able to wring out an extra 9 to 12 dB of gain. I'm pretty aggressive with the envelope tool and the applause between tracks doesn't always sound 100% natural, but for 99%  of my listening purposes it doesn't matter all that much to me.

[jj69]

Thanks for the responses.  I'm looking at the envelope tool in Audition.  It's basically a graph and you draw the shape of the curve you want with it. 

Any recommended settings for audience breaks? 

[vanark]

You will need to try it out to see how it sounds. I've found you can be more aggressive on the beg. and end than you think.

[nulldogmas]

I'd say look at the background level of audience noise that you don't want to reduce (i.e., everything except Mr. WHOOO) and start bending the curve just above that point. Trial and error should get you to the best possible result, but even a 3:1 or 4:1 slope shouldn't sound too unnatural.

[jj69]

Is messing with the volume the only way to accomplish this result, or is there another way?  I was expecting the suggestions to involve some sort of compression. 

[Brian Skalinder]

Is messing with the volume the only way to accomplish this result, or is there another way?  I was expecting the suggestions to involve some sort of compression. 

Nulldogmas comments relate to using compression.  FWIW, my general experience is that using hard limiting on clapping sounds tolerable, but on music and especially vocals it more often than not sounds terrible.  I'd try compression, starting to compress right around the "good" content levels you want to keep, and using a strong-ish ratio -- as nulldogma suggests.  As nd also indicates, the exact ratio will take some trial and error to get it sounding best to your ears.

[hoserama]

It's time consuming, but you could use the Spot Healing Brush tool in the Frequency Spectral display of Adobe Audition. Claps are usually very isolated frequencies somewhere between 1-3k, and you can easily zap those out and nobody is the wiser. For the "wooo", you can usually see the frequency band it resides in...just use the spot healing brush and wipe it out. It's like an audio frequency airbrush.

If you throw up a sample I'll give it a quick take.

[jj69]

I've used the spot healing brush.  I've found that for claps, whenever possible the clip/pop remover is the best way to remove them with no artifacts.  I only use the brush when the clap is buried among other sounds and the only way to find them is in the spectral view. 

However, I have found that if I use the brush to remove other sounds, it often leaves behind a feint buzzing artifact that is not pleasant.  It's possible I don't have the settings right, but I've found that it's easy to get too heavy handed with the brush. 

Most of the places where this guy is screaming are between songs, or during quiet passages following a solo.  If I use the brush to remove it, I think I'll end up with some pretty bad artifacts. 

[hoserama]

I guess I haven't gotten as many artifacts. When I didn't like how Audition handled it, Izotope RX2 could also do something very similar to it with often different results.

[nulldogmas]

Nulldogmas comments relate to using compression.  FWIW, my general experience is that using hard limiting on clapping sounds tolerable, but on music and especially vocals it more often than not sounds terrible.  I'd try compression, starting to compress right around the "good" content levels you want to keep, and using a strong-ish ratio -- as nulldogma suggests.  As nd also indicates, the exact ratio will take some trial and error to get it sounding best to your ears.

Yes, sorry, I was referring to a dynamic compression graph. Though looking more carefully now, it appears that the envelope tool in Audition (which I don't have) may work differently, in which case my advice wasn't very helpful.

The two programs I've used the most for this are: Cool Edit Pro, which give you a little input/output X/Y curve that you can manipulate, and where you want to start curving the line to almost flatten out horizontally, starting at the point where only the loud guy will be affected; and SoundStudio, which lets you pick a dB level (which would be that same level-that-only-loud-guy-hits) above which you indicate how much compression to apply (3:1 or 4:1 is usually fine for things like audience noise). Different interfaces, but the general principle is the same: You want to deloudenize (unenlouden?) everything above a certain point.