Automating DeEsser +Alternative DeEssing Schemes

[ojan327]

I'm using Compressor as a DeEsser and I like it compared to the 'DeEsser' plug-in.

 

However, I wanted to know if I'm crazy or are there other people out there who automate the DeEsser threshold on vocals to try to get them as close to perfect as possible?? (In my case, not only controlling S's, but Z's, CH's, P's, V's, etc.)

 

Any feedback on this topic will be greatly appreciated!

[David Nahmani]

I wanted to know if I'm crazy or are there other people out there who automate the DeEsser threshold on vocals to try to get them as close to perfect as possible?? (In my case, not only controlling S's, but Z's, CH's, P's, V's, etc.)

Whatever works I guess, but you may ask yourself if it wouldn't be faster to simply automate your volume using track automation, and not use any plug-ins.

[ojan327]

Thanks for your reply, David.

 

Automating volume does not produce the same result. By automating the DeEsser threshold, the volume consistency is maintained while the unwanted rough sound is suppressed.

 

It's working for me, it's just really taking a lot of time due to specific thresholds I have to set for every occurrence of sibilance that needs processing.

[Holger Lagerfeldt]

You can do as David suggests, but you need to be very precise with your automation. If you're too fast or too late you'll change parts of the vocal that weren't supposed to be changed.

 

But the same could be said about automating the threshold in the compressor/de-esser. You need to be very precise.

 

Another solution is to set the threshold so only sibilant areas are affected. The threshold stays where it is. You then automate the ratio only. This way you will never risk de-essing too much or too fast/late, but you can control how much or how little.

 

The downside of the latter process is that you won't be able to de-ess on sibilants that are lower than the threshold.

 

A third solution is to make a duplicate of your vocal track, including all processing, but insert a Utility > Gain plug-in at the end of the chain and flip the polarity (called phase invert in the Gain plug-in). Then use the scissors or marquee tool to remove everything but the sibilant areas. Use volume automation on the remaining regions to control how much de-essing you want.

 

The theory behind the above method is that you're simply phasing the sibilance out. This method does not require any de-esser plug-in in the chain. This is the method used by the Eiosis de-esser plug-in.

 

None of these methods are particularly fast, but there's no way around manually addressing major sibilance problems. The real solution is a good vocalist, a good microphone and vocal chain/processing chain. Using a very fast attack in your vocal compressor is generally a good way of catching transients, i.e. 1.5 - 2 ms.

[fader8]

A third solution is to make a duplicate of your vocal track, including all processing, but insert a Utility > Gain plug-in at the end of the chain and flip the polarity (called phase invert in the Gain plug-in). Then use the scissors or marquee tool to remove everything but the sibilant areas. Use volume automation on the remaining regions to control how much de-essing you want.

 

The theory behind the above method is that you're simply phasing the sibilance out. This method does not require any de-esser plug-in in the chain.

ojan,

This particular method lagerfeldt describes is the most transparent and I use it for a lot of restoration work. Here's an enhancement to this technique that improves on it further, and avoids most of the editing...

 

After duplicating the track and reversing its polarity, place a Linear Phase EQ, followed by a Noise Gate, in the last inserts of the dup track. Solo this track. Bypass the gate for the moment.

 

Set the EQ with a bandpass or highpass around the sibilance. Keep the slope gentle, don't brickwall it or there'll be too many artifacts. You want to hear the sibilance nice and loud but not worry about eliminating everything else.

 

Enable the gate. Set Attack, Hold and Release to 0 and add 3 or 4 mSec of lookahead. Start by adjusting only the Threshold, Reduction and Hysteresis and try to isolate the sibilance completely. Use the Side Chain filter if the gate is still triggering on other frequencies you want to keep.

 

Now add the original track to the mix to allow the gated sibilance to cancel the original. You can tweak the gate's Attack, Hold and Release if needed but you shouldn't have to. Just be careful as you can add nasties to the sound with weird settings.

 

The key benefit to this is that you won't be reducing the level of lower vowel frequencies that may be concurrent with the sibilance. It avoids a lot of region editing too!

 

edit: Note that you can always tweak the linear phase eq for where and how much sibilance gets removed, but remember that you don't want to remove too much of it as you need some to keep the vocals intelligibility strong in the mix.

Edited December 3, 2011 by fader8

[Holger Lagerfeldt]

Nice addition to the technique, I think you posted this before but I forgot all about it!

[camillo jr]

fader8, I'm gonna have to try that. I've done the "phase reverse track with esses" technique before and it makes for excellent results but it's pretty work intensive.

 

BTW, did you mean high pass filter rather than low pass for the Linear Phase EQ?

[fader8]

BTW, did you mean high pass filter rather than low pass for the Linear Phase EQ?

Yes, sorry, morning brainfart. I'll fix it. I want to add a note to it anyway.

[fader8]

I think you posted this before but I forgot all about it!

I'm sure I have at some point, maybe more than once. I'll move this to the Tips&Tricks forum. That'll keep me from having to repeat myself without realizing it! Kind of an anti-dementia thing.

 

Pursuant to this thread, I notice a lot of mixers over-de-essing. More so with hiphop for some reason. There's a syndrome that happens when you listen to a song too many times while mixing. You learn the lyrics so well that you can mangle the heck out of them and still believe that you can actually understand what's being sung. And you can, but that's only because you already know what they're singing! So if you have to do a lot of de-essing on a bad track, have someone with fresh ears give it a listen and see if they can understand the lyrics. You might be surprised at what they'll tell you they can't hear!

[camillo jr]

OK, just did some experimentation here (before seeing your second-last post), trying both the low shelf with a cut and then the high shelf with a boost. I found that it was actually easier to get really good separation of the esses on the dup track by boosting the Linear Phase EQ's high shelf. And engaging the noise gate's side chain filter was also necessary.

 

After setting everything up, it was very easy to adjust the amount of de-essing simply by adjusting the gain plug's level, which I had parked around -9db. On the LP EQ, I was surprised by how gentle the slope and the boost was to make this work.

[fader8]

After setting everything up, it was very easy to adjust the amount of de-essing simply by adjusting the gain plug's level, which I had parked around -9db. On the LP EQ, I was surprised by how gentle the slope and the boost was to make this work.

A lot of settings are going to be content dependent/specific. What you're showing looks like it works well for esses, but if you're also dealing with plosives, fricatives, etc., it might not cut it. The nice thing about this is you can just set up as many duplicate tracks as you need to deal exclusively with those!

 

As for actually boosting with a shelf like that instead of just highpassing, I've often noticed, (for example with denoising) that sometimes it's better to exaggerate the error first before removing it. Things just seem to get a better "bite" I guess!

[jschmidtiii]

Hi all,

 

This forum is incredible. Fader, quick question about what you said here:

 

After duplicating the track and reversing its polarity, place a Linear Phase EQ, followed by a Noise Gate, in the last inserts of the dup track. Solo this track. Bypass the gate for the moment.

 

Why linear phase EQ vs. channel EQ? I don't have a firm grasp of the difference between them. I did some searching and learned that linear phase is a bit more cpu intensive and is generally used for mastering, but that's as far as I got. Thanks!

[fader8]

This forum is incredible. Fader, quick question about what you said here:

 

After duplicating the track and reversing its polarity, place a Linear Phase EQ, followed by a Noise Gate, in the last inserts of the dup track. Solo this track. Bypass the gate for the moment.

 

Why linear phase EQ vs. channel EQ? I don't have a firm grasp of the difference between them. I did some searching and learned that linear phase is a bit more cpu intensive and is generally used for mastering, but that's as far as I got. Thanks!

For a signal to effectively cancel out another signal, the phase relationship between them needs to be dead on, particularly when working with high frequency content like this. Nonlinear phase filters, like the channel eq, or any other, not only shift the phase of the original signal, they shift it differently for different frequencies. So there's no way to line them up to effectively cancel.