Help with de-essing vocals?

[RemoHead]

Hi guys,

I wonder if anyone can offer me some advice with regards to de-essing vocals, I have a couple of different de-essers, namely the Waves ren de-esser and the UAD precision de-esser.

Im currently recording my vocals through an audio technica AT4033 mic into a GAP-73 or a Focusrite voice master, which in turn feeds my mac running logic 8.

Any tips or tricks you could offer would be greatly appreciated as this is a subject area that I could do with a touch of help with!

thanks,

Chris

[Hugh Robjohns]

First, optimise the vocal performance and then the mic position to minimise the sibilance in the first place.

Record with plenty of headroom -- sibilance is made much worse if the analogue circuitry and converters are running close to their headroom limits. I'd suggest peaks no higher than -10dBFS when recording.

After that, it comes down to careful optimisation of the de-essing tools you have. Experiment with the frequency setting (if provided) and only do just enough compression to fix the problem.

hugh

[Matt Houghton]

...and for a more detailed discussion, you could check out Mike Senior's recent article in SOS.

[Henry-S]

I found this very hard to sort out and one small piece of advice from me that seemed to work really well was making sure I used my ears and not my eyes. I'm pretty sure the Waves De-Esser has a "side chain" so you can hear how much you are actually cutting from the signal which is a great thing to check. I would sit down and do this

Listen to the vocal and then setup the De-Esser and use an EQ to find exactly where the S's were most present. I then would look at the threshhold and then click the mouse and carry on listening and close my eyes. I would then slowly drag down the threshold without looking at the screen and use my ears to tell me when I felt it had gotten rid of the S's which were too loud. Remember, people still make S sounds when singing and if you nail them down too much you can give your singer a lisp (very easy to do when just using your eyes!).

I would then listen to side chain and check that I wasn't taking out too much from the vocal. If there were bits where it was not taking enough out or taking too much then would automate these parts via the threshhold.

[Doublehelix]

The Precision De-Esser from Universal Audio also has a "listen" function where you can hear what is being taken away.

I have found that you can reduce some of the worst of the sibilance, but do not expect miracles without damaging the rest of the voice.

Certainly, Hugh's advice about getting it right at the source is the best way to start. If subtle de-essing is not enough, you can resort to waveform automation, which is a bit of a pain, but fairly effective if done right. Make sure you zoom in so you are only pulling down the esses.

[BenLD]

I have the Precision de-esser - and yes the "listen" function is the solo [s] button under the frequency dial

Hit the solo button, turn the frequency knob to where you can best hear the problem frequency (and play with the width knob to really home in on a narrow frequency band if necessary); then turn the threshold knob down from zero until the compression starts to kick in (the gain reduction meter will show you how much is being applied); then hit 'solo' again to hear the effect in context

The other feature I really like is the 'split' button - it defaults to 'off', which means the whole signal is being compressed when the de-esser is triggered (this is the way 'classic' de-essing has always worked). But if you switch it to 'on' then only the offending frequency is compressed and the rest of the signal is left untouched - which is a much more subtle effect IMO.

I don't have the Waves de-esser so can't help you there!

Cheers

Ben

[mooblie]

Hi - as a videographer, who lurks here for sonic expertise (and only occasionally posts) can I ask:

What exactly causes sibilance when recorded? I sense it's not a simple as "clipped high frequencies"??

I use wireless microphones (near lecterns for readers at weddings, etc.) and I'm often faced with sibilant recordings, and I never quite know what caused it, and the best way to deal with it.

Using Sennheiser G2 wireless gear (with HDX noise reduction), recorded to HDV cameras (which record audio in MPEG-1 , 16bit, 48kHz format - recorded levels kept around -20dBFS, monitored on phones, but it's live.... and one-take-only stuff... and there's the images to think about.....).

TIA for any insight from the experts here!

[Hugh Robjohns]

In the case of radio mics, high input levels can lead to emphasised sibilance, especially if the internal limiter is triggered, largely because the allowed TX bandwidth is restricted.

Sibilance is a natural part of most speech, but it generates an unusual amount of strong HF spectral components and that can stress analogue circuitry in a way which isn't always obvious.

Most budget equipment has significantly less headroom at HF to lower frequencies, so it's not that hard to find sibilance pushing the envelope beyond the comfortable zone for the equipment, and that results in even more distortion which makes the sibilance sound even worse.

Backing off the input level usually helps in these situations.

hugh

[mooblie]

Thanks Hugh - I will experiment. I have certainly had more trouble with:

(i) a (higher output) Sennheiser K6/ME64 feeding an SKP500, than

(ii) a (lower output) Sennheiser E914 feeding the same SKP500.

The suggestion is a sibilant speaker is maybe overloading the transmitter? (The receiver/recording levels are nowhere near overload.)

[RemoHead]

Thanks guys for all your suggestions, i'll def give them a try, ive been experimenting with the couple of de-essers ive got with mixed results! lol, thanks again.

[Evie McCreevie]

De-essers are crap. Time-consuming crap too - you can waste ages fiddling with them. Some settings work on some parts, then fail on others.

FWIW - my solution...

Stick a de-esser on temporarily when working on a track, just to save your ears. When the vocal is finally comped and ready, take off the de-esser.

Go through the track and manually drop the 'esses', either via automation or hard edits. Make a safety if you're scared.

It only takes 5 or 10 minutes once you have the hang of it.