If the volume on your copy of the audio is lower you’ll have to set a lower threshold on the de-esser to achieve the same level of de-essing, say -36dB instead of -33dB,
[ “Maximum attenuation” should be 5dB-10dB lower than threshold, & leave all the other de-esser settings as they are ].
I would add RMS normalize immediately before De-ess , because De-ess is a (volume) threshold-dependent effect.
Then you can keep the De-esser threshold setting constant for your whole project.
RMS normalize is a more reliable way of creating consistent volume across your project.
Audacity’s native normalize tool goes by peak-volume values, not RMS values,
so one momentary loud peak in a recording can knock the whole thing out-of-whack.
That’s interesting, thanks. I had been normalizing to the default of -1 dB, based on my reading that you need to leave a little headroom to avoid clipping for later effects. RMS normalize has a default of -18 dB, which seems a lot. Would an RMS normalize of -1 dB still be reasonable?
Let’s try this again. Here is the result. I’m probably too close to it now - half of me says it sounds ok, half says there are still deficiencies. Maybe de-essing more? Does it sound muffled - tempo sped up too much? Any advice appreciated.
Sorry, that one had the pseudo stereo track at -3dB, here it is at -12 dB. Hard to keep track with seven effects. This one is better. But same questions - more de-essing? Does it sounds mufflled, result of something being applied too much?
Aha, it is caused by the mixing together of the two tracks - mono and pseudo stereo.
I tried several things. Reducing the RMS normalize step to -24 dB. Adding an RMS normalize and limiter again to the end of the chain. Also found that Ozone Imager created a little bit of clipping, so added a limiter after that. However, no matter what, when mixing the mono and pseudo-stereo, even though no clipping in either, resulted in a track with a lot of clipping.
The only thing I found worked was using plain normalization, which does not increase it nearly as much. Once mixed, the 11 minute audio segment then only had one clipping place, and when I listened to that section, could not really hear any distortion. Unfortunately this does not make it nearly as loud. So I had two thoughts, neither of which seems great, comments very welcome:
—a. I could create it this way with plain normalization, import it into my video editing software Camtasia, and then increase the volume with the simple volume up control from there. But based on my understanding of your comments, the plain normalization with much less increase than using RMS will not result in as good de-essing. Might compromise the presence boost as well?
—b. Create it this way, then RMS normalize the final result. Tried this, and it creates a lot of clipping, so then added a limiter. Again, since the de-essing was earlier in the chain after plain normalization, probably not getting as good results?
Thanks again. I think this is the best I can do, not bad I think, certainly much better than the original. One question, I’m wondering if I am oversensitive to the s’ing now, or does this still need more? If more, would you be so kind as to suggest slightly more aggressive settings than the “reasonable” ones you mentioned earlier?
Back to the nitpicking zone, (you did ask).
There are still some clicks which eluded the de-clicker, made more-obvious by the presence treble-boost. They can be removed manually (individually) .
But manually removing them is probably too time-consuming if you have hours of audio to process.
Thanks. I’m pretty much there, and the first full track I did sounds good. The second though seems to have some fluctuating volume changes. To my imperfect ear it gets breathy and then quiet more than once within this short sample. For example, can you hear how “project” seems to get quieter/muffled?
I have included the original for comparison, which does not seem to have this, so am wondering if it was my processing. The processing order at this point is: increase tempo, decrease pitch, normalize, ozone imager, mix, boost eq, de-ess.
On that there’s an occasional metallic rattle in the 8kHz-14kHz range that will have been exaggerated by the treble-boost, (and possibly made worse by other processing, like pitch-shift).
There wasn’t any of that on the examples you’ve posted previously.
The rattle can be removed using Paul-L’s precision de-esser, (even though it’s not sibilance) …
But doing that will reduce the treble/presence,
Update: the metallic-rattle can is artifact which can be created by Audacity’s “high quality” pitch-shift, & “high quality” change-tempo. So if you must pitch-shift or change tempo, don’t check the “high quality” tickbox on those effects.
By checking it at every step along the way, the “fluctuation” appears to be caused by the final de-essing step. I’m using the “reasonable” settings as suggested. It does not appear to have this effect on other tracks I have processed so far, so I’m a bit baffled. Can you take a listen to the before and after attached? The first segment has all the processing in, through boost presence. The second segment just has the addition of the de-essing. At least to my ears the second segment is going up and down, or in and out, or or muffled and clear, as you wish, several times within that short segment.
I think de-essing is necessary, but the result on this track sounds disorienting, at least to me. Any thoughts?
Actually, on checking, it is happenning on all the tracks. Can you take a listen to the attached? Notice how the word “essential” in the second segment, after the de-essing, sounds like it is behind a blanket. Perhaps just use much less de-essing? If so, do you have recommended “less than reasonable” settings