sibilance from 3k to 16k ??? (Audio inside)

[ches1]

hi everyone

I'm having a lot of trouble to eliminate some sibilance in a vocal track.
I tried a frequency sweep with both a narrow and wide q but I coudn't find THE freq. that was problematic, it sound like most of the highs are contributing to the sibilance.I tried some de-esser but unsuccessfuly.
the only way I was able to eliminate all of it was to low pass at 3.8k, needless to say it now sound VERY bad.

if yall could take a listen to this and give me advice it would be greatly appreciated.

P.S. I am not the rapper nor did I do the music so I can't re-record anything.

 

[dementedchord]

in most cases this is done by useing a very narrow Q... boost the hell out of it... and sweep looking for the nasties... then pull that range out playing around abit with the Q and how much you decrease to soften the effect...

 

[ches1]

like I said I already tried that, is it common to make 2-3 different cut to solve a sibilance problem(as opposed to find out that 6k is the prob and reducing only this)???

also, should I eq/de-ess before or after compression???

 

[SouthSIDE Glen]

like I said I already tried that, is it common to make 2-3 different cut to solve a sibilance problem(as opposed to find out that 6k is the prob and reducing only this)???

also, should I eq/de-ess before or after compression???

I'm about to walk out the door and won't be able to analyze your file until later tonight - if you still need it by then. But some quickies in the mean time:

It is not uncommon to have to surgically target for than one frequency, though that doesn't happen as often in the high freqs as it does in the upper mids. Go ahead and remove what jumps out even if 3 or 4 frqequs jump out. I also might recommend to be on the safe side to start your sweep low, maybe around 2.5-3k or so; you might have some troublesome upper mids contributing to a harshness that is not exactly sibilance.

Also, take a look at a FFT frequency analysis and look to see if you might be getting some harmonic problems in there. For example, you might have a problem bump at 3k, foloowid by bumps (perhaps smaller, but recognizable nonetheless) at 6k 12k and 18k. Those number are just an example, the key is to look for a regular mathematical pattern. If you find that, you can speed things up by applying a harmonic filter to it instead of hitting the frequencies one-by-one.

I'd clean up before compression and then re-sweep after compression to clean up any leftovers.

G.

 

[Sonixx]

post only the vocal...

 

[SouthSIDE Glen]

OK, I checked you sibtest sample, and I have an immediate question: are you having the same problem with your master WAV file or is it just on the MP3?

I ask that because what I personally am hearing and seeing on your MP3 is not "standard" sibilance (if there is such a thing) but what appears rather to be artifacted sibilance actually happening largely above 15k. It sticks out like a sore thumb above the typical MP3 rolloff above 15k, and corresponds quite well to the sibilant and percussive parts of your vocal (see attached spectral time display; the peaks above the wall at about 15k correspond to the sibilant points in the program.)


I can't speak to the actual WAV, but on the MP3 a simple low pass at 10k with a Q of about 7 using RND's Unequalizer low pass function seemed to smooth out the sound fairly nicely.

G.

 

[Tim Gillett]

The vocalist sounds to me to have a "sibilant" voice, meaning the frequency is higher than usual for a male adult. He possibly hangs onto his sibilants a bit longer than usual. Would this be regarded as a problem if the vocalist were female?

 

[ches1]

post only the vocal...

I can't speak to the actual WAV,

here it is

Glen, Thank you for your time.
Ima re-try a sweep and make a few cut to see what happens
I'll also try to look into that FFT and Harmonics filter thing too, I think I saw that in Audition.
when you say "the peaks above the wall at about 15k" do you mean every peak 15k and up that are clearly crossing the thick rectangle??

Peace.

 

[SouthSIDE Glen]

when you say "the peaks above the wall at about 15k" do you mean every peak 15k and up that are clearly crossing the thick rectangle??

Pretty much, yeah. There may be a small occasion here or there that isn't directly related, but most of what you see in that spectragram that rises above about 15.5k or so corresponds to a sibilant or percussive moment in the vocal track.

If you look at the FFT (which I did also) you probably won't see much that's immediately recognizable as untoward sibilance in the "typical" range (around 6-8kHz); the area is filled with peaks and valleys, but nothing really jumps out at those key points in the vocals. This matches the spectragram which also shows no noticeable concentration of energy in that frequency range at those points in time (or anywhere else in the clip, FTM.) More importantly, it matches what I hear, which isn't the typical harsh, fuzzy-sounding sibilance, but rather something with more of a colder high frequency sizzle.

Back on the FFT, you can visually confirm this by looking up at about 16k and above. As with most MP3s, the FFT slopes down sharply and crashes to zero right around 15k or slightly above. But when the sibilant or percussive parts of the vocal hit, you'll see the area just to the right jump with some smaller but significant lobes of energy.

Once again, I gotta ask you what YOU hear. Are you hearing this on the original WAV of the vocal itself, or do you not hear it until you sum your tracks into a stereo mix, or is it in the MP3 encoding that it appears?

EDIT: OK, I just took a listen and look at the vocal-only WAV file. It is extremely rich in high frequency, all the way up to about 22k. At the sibilant points there does seem to be some peaking around 8-10k - which in itself is on the high end of typical sibilance, but even more striking IMHO is the amount of energy *above* 10k at those points. Simply put, that is an extremely bright recording.

Attached is an FFT snapshot taken at about 24.9 seconds into the file. Note the wide "peak" extending from about 8k all the way up to about 14k, and the fact that though things drop off from there, there is still quite a bit of energy all the way up to 22k (the end of the chart).

I'm not sure of the cause at this time, but I'd still start with trying what I suggested on the MP3, except on the vocal WAV itself, and that is to low pass at about 10k or so with a fairly steep slope to take out a lot of that HF energy.

G.

 

[ches1]

I hear it on the original WAV 100% dry. I did compress on the way in but nothing more than about 2db of gain reduction. the audio I posted are further compressed but it did not accentuate the sibilance.

I don't fully understand your last advice because I don't know much about FFT, but I'ma do my homework and then check it out

 

[SouthSIDE Glen]

I hear it on the original WAV 100% dry. I did compress on the way in but nothing more than about 2db of gain reduction. the audio I posted are further compressed but it did not accentuate the sibilance.

I don't fully understand your last advice because I don't know much about FFT, but I'ma do my homework and then check it out

Oops, I just added an edit to my last post and didn't see your follow-up post. Sorry about that. Go back one step and read my edit at the end of the last post.

G.

 

[ches1]

thanks glen, I'll go now and try all that

 

[Sonixx]

I checked your vox clip and isolated a small sibilance section to determine the energy distribution. it's fairly wideband for sibilance.

to combat there are several approaches...

- manually reduce each by hand. for only a song or two it's not that big of a chore but for a CD's worth it can be tedious.

- boost/isolate the frequency range and drive a compressors side chain with fast attack and release.

IMO the sibilance is not that exaggerated and won't take much.

 

[ches1]

soooo, I did the 10k lowpass thing and resweeped before and after compression and cutted respectivly. it worked ok. the vocal losted a lot of it's sibilance AND clarity, now it sound somewhat muddy.
so I'm asking myself what's worse, sibilance or muddyness???

should I use a de-esser instead of EQ?
I didn't because the only de-esser I have is not doing it good as far as I can tell but it seems like it's a more logical approach to reduce sib. because it's only taming the "wrong" freq. when they occur, leaving the rest of the signal untouched when it's fine.
what y'all think is the best de-esser plugin available??

Simply put, that is an extremely bright recording.

how did I ended up with such a recording???
I used an M-audio Luna LDC (wich is considered "warm" by some mag.) through a DMP3 then a compressor(lightly) than in the pc with a good soundcard.
can it be the mic or the compressor or is it totally the rapper fault?

 

[SouthSIDE Glen]

soooo, I did the 10k lowpass thing and resweeped before and after compression and cutted respectivly. it worked ok. the vocal losted a lot of it's sibilance AND clarity, now it sound somewhat muddy.

It sounds to me like you may have done too much in one chunk. Use your ears as you go along, and try one step at a time. When I recommended the low pass, that meant performing just the low pass and nothing else. I'd try that by itself first. If that does't do it for you, then skip the low pass and try the sweeps. If it still sounds kinda sharp or brittle to you after that, try adding some low pass at maybe around 13k. The main point is you gotta let your ears tell you how far to go.

should I use a de-esser instead of EQ?
I didn't because the only de-esser I have is not doing it good as far as I can tell but it seems like it's a more logical approach to reduce sib. because it's only taming the "wrong" freq.

It never hurts to try it for yourself, ches; as long as you keep your original file, you can always go back and never permanently hurt anything.

That said, though, I have my doubts about a de-esser, because what you have is not standard, textbook sibilance; your high frequencies are "sibilant" over a good three times the frequency range of "normal" sibilance, and your high frequencies remain too rich all the way up to the end of the spectrum. But go ahead and try it if you like, maybe it'll suit your needs.

The main reason I recommend the low pass (and I tried it myself) is because, frankly, on that vocal track there's nothing really happening at those high frequencies other than that sibilance; the human voice just does not naturally go that high for much else. If 10k is too much for you you can always adjust the cutoff frequency or the steepness of the slope to fine-tune it. But anything more complicated than that should not be necessary; as Sonixx said, the problem is really not that bad and - on my desk anyway - it is an easy fix.

How did I ended up with such a recording???
I used an M-audio Luna LDC (wich is considered "warm" by some mag.) through a DMP3 then a compressor(lightly) than in the pc with a good soundcard.
can it be the mic or the compressor or is it totally the rapper fault?

It's probably at least partially the rapper; a combination of a naturally sibilant voice with his perhaps working the mic too tight and too on-axis.

I am not familiar with that particular mic, so I can't speak to that directly; though I would not expect ultra-high quality from budget gear to begin with.

But you know what; your other synth tracks on that song sound awfully bright as well. You have something going on there that's putting a crackle on almost everything in that song it seems.

G.