Ok, where should it be put in the chain? Chain is currently: change tempo, change pitch, normalize, ozone pseudo stereo, presence boost, de-ess.
Whatever effects were applied here, rattle-reduction EQ should be inserted after that.
[definitively not before “change pitch”].
If that’s Audacity’s native normalize, rather than RMS normalize, you’re unlikely to get consistent volume.
Without consistent volume, de-essing will be inconsistent, (too much if it’s loud , too little if its quiet)
De-essers only act above a volume threshold, to get consistent de-essing the audio it’s being applied to has to be of a consistent volume.
Ok, thanks, I’ll put rattle reduction at the end of the chain.
Re RMS normalize, I found it created so much clipping, however I’ve played some more, and found that -22 dB produces a track significantly louder than peak normalization, and with just a bit of clipping, so seems to be a sweetspot for my tracks.
However, Ozone Imager then adds a bit more clipping, and mixing that with the mono creates more clipping again. So the question seems to be when to apply the limiter. Doing it after RMS normalize doesn’t seem right, because I’d have to do it again after the Ozone and mix.
Therefore, should I proceed with (a) change tempo, (b) change pitch, (c) RMS normalize to -22 dB, (d) Ozone, (e) mix, and then (f) apply the limiter? And then (g) EQ boost, (h) de-ess, and (i) rattle remove?
The attached is the result of that chain.
If, immediately after “RMS normalize to -22 dB”, you limit to -6dB rather than -1dB, that should give you more headroom : and reduce the need for limiting again at subsequent stages.
Thanks again. Result attached. Anything else?
That sounds too sibilant to me.
The pattern of sibilance is now too complicated to be corrected by Steve’s de-esser, which just has 1 band.
Paul-L’s De-Esser can have dozens of bands, 5 will do …
Paul-L’s De-Esser should only be used on mono, as it will add weird effect on stereo/pseudostereo.
Rather than use two de-essers in your chain, just use precision one : Paul-L’s.
(a) change tempo,
(b) change pitch,
(c) RMS normalize to -22 dB
(d) (soft) Limit -6dB
(e) de-ess using Paul-L’s
(f) rattle remove EQ
===================== below are deluxe options
(g) Ozone (pseudo-stereo)
(h) EQ boost (presence)
Ok. This means I need to reorder the chain, as much of it had been done after the pseudo stereo. I’m not sure how the effects affect each other. Does the following make sense?
-
Change tempo
-
Change pitch
-
RMS normalize
-
Limiter
-
EQ remove rattle ← should this be done before normalize, even first?
-
Boost EQ
-
Paul de-ess
-
Ozone and mix to pseudo stereo
-
Limit again if (8) introduced clipping.
I think the rattle-reduction EQ was based on audio you had pitch-shifted, if so then it must come after “change pitch”.
IMO De-essing should come before “Boost EQ” (i.e. before treble-boost/presence).
The final step should be RMS normalize, then limiter, (not just limiter).
Then the output of your chain will have a consistent volume from one batch to the next.
Ok, here’s the order I applied:
- Change tempo
- Change pitch
- EQ remove rattle
- Paul de-ess
- Boost EQ
- Ozone and mix to pseudo stereo
- RMS normalize
- Limit
(I will post all the settings in one consolidated message for the final version.)
Result attached. Better? Any improvements required?
That sounds quite good, no rattle on that one, but needs more de-ess IMO
More frequency bands will do it , but that means longer processing-time …
alternatively leave the settings on Paul-L as is, & add-on Steve’s DeEsser, (on “reasonable” settings), to the existing chain, which will be quicker than increasing from 5 to 20 bands on Paul-L, but the results are less precise …
Yes, moving from 5 to 20 bands increases the time significantly. But in for a penny, in for a pound. Might as well do it right.
Here is the result. Any better? Any additional improvements you recommend?
Almost perfect , but personally I’d lower the de-esser threshold a bit more : by 1 or 2.
When I apply the Paul-L de-esser settings above to that it still manages to shave off a noticeable amount of sibilance.
I’m not suggesting you use the Paul-L de-esser twice, but that you use it once with a slightly lower threshold so it removes more.
Basically lower the de-esser threshold until you’ve gone too far, then back it up by say 3dB from the “too far” setting.
That’s how I set the de-esser threshold.
Ok. Does “lower the threshold” mean lowering the threshold setting from -26 dB in your previous post to -27 or -28, as the only setting to change?
Yes : try -27, -28, -29 … until the sibilance is not intrusively loud.
Note the upper frequency threshold is 10000Hz , at one point it was set at 14000Hz.
Applying rattle-reduction equalization means there is no excess sibilance above 10000Hz,
I’m pretty sure I don’t have the sensitivity to these frequencies to make a good assessment. So I tried -28 in audio-excerpt1, and -29 in audio-excerpt2 attached. Do either of these sound better, or do you think more is required?
Both -28, -29 are quite acceptable, but I would still apply a bit more DeEss.
If I apply -26 to your -29 it still manages to shave-off more sibilance …
The hardware & wetware you have will determine the extent of the upper frequencies you can hear.
Ok, thanks. Running the de-esser twice would mean many many hours processing, which I would like to avoid since there is a fair amount of video work to do as well. Is there a formula or way to determine what applying -26 to the -29 amounts to, such as just do -30 once?
No formula , just trial & error : applying the Paul-L’s DeEsser once, but at lower thresholds : -30, -31, -32 …
Ok, I tried several options, and think it is one of the attached: first is -31 dB, second -32 dB, third -33 dB.
Since you heard -29, if you think -31 is too much, it must be -30. And if you think -33 is not enough, it must be -34!
What say you?
And anything else you think it needs?
-33 is very close to the settings I would chose.
The difference is small, but audible to me , & visible …
-34 looks like the magic number.