Popping P's

[John F]

Does anyone know of a remedy within SAWStudio for popping P's? I know I can highlight a section of a soundfile within the multitrack and lower its volume, but that doesn't seem to give me the precision I'm looking for. On the other extreme, I could also zoom way in to the pixel level and drag the pixels down, but that seems like a lot of extra work.

My current cure is to open up that soundfile in Adobe Audition and use its "Autoheal" function, which works very nicely - but it would be nice to stay within SAWStudio if I could. Thanks for your consideration.

John Francis
Rolla, MO

[MMP]

I often use a soft-edge for this by removing the front half of the P and starting a soft-edge at the beginning, or automate the high pass filter to remove energy below 120, or both. Izotope also makes a deplosive plugin that works well.

[John F]

I often use a soft-edge for this by removing the front half of the P and starting a soft-edge at the beginning, or automate the high pass filter to remove energy below 120, or both. Izotope also makes a deplosive plugin the works well.

Both great suggestions - thank you, Michael!

John Francis
Rolla, MO

[jmh]

If you have levelizer, you can:

Duplicate your vocal track with levelizer patched in the prefader fx,
use the (levelizer) gate with eq high cut,
Remove Silence automatic (in levelizer),
paste a silent region to the right of the last plosive (this is to let the ducking recover *),
when you are confident that you only have 'P's, select that track (so it is the only one playing) and build mix mono to a hottrack layer,

You can now use the new layer (mute it - it will still function as a key) as a key source for gating with reverse engaged on the original vocal and use floor to dial in the exact amount of attenuation you want. You might slide the key audio to the left about as many ms as your attack to get a smooth action. Optionally you can use the compressor rather than gate to attenuate the P's more proportionally.

This method takes a bit of time, but if there are enough plosives to justify doing it, it can give great accuracy - as other than the plosives, your key track will be all silence.

John

* Having contiguous audio for the key avoids gotchas related to Saw's keying if there are gaps.

[Todd R]

I use Acon Digital Restoration Suite's DeClicker VST which has a plosive reducer. I find it works well for our run and gun lavalier mics....but it not so good on the large diaphragm condensers in quiet VO situations

Even with pop screen (homemade dbl panty hose hoop), Acon DeClicker plosive reducer is not enough...I can see the plosives in Fab Filter ProQ3 at around 70Hz, but haven't found a solution to key reduction fx...whatever dynamic processing I apply is more detrimental to overall tone than whatever benefit I might see in applying reduction

The DeEsser VST I usually use (Fab Filter ProDS) works great tho...but clearly not for plosives...

Acon Digital Extract Dialog can be a lifesaver for some stuff...but also it's not great on plosives into the condensers I'm using (Rode K2)

sadly, I usually look for the best performances rather than try to fix in post

BTW I despise overall compression and proximity effect on VO. I only ever use dynamic EQ on troublesome freqs depending upon the contributor

[jmh]

sadly, I usually look for the best performances rather than try to fix in post

Yea, it's truly sad when you capture those tracks that leave you with no challenges to overcome...

[jmh]

If you have levelizer, you can:

Duplicate your vocal track with levelizer patched in the prefader fx,
use the (levelizer) gate with eq high cut,
Remove Silence automatic (in levelizer),
paste a silent region to the right of the last plosive (this is to let the ducking recover *),
when you are confident that you only have 'P's, select that track (so it is the only one playing) and build mix mono to a hottrack layer,

You can now use the new layer (mute it - it will still function as a key) as a key source for gating with reverse engaged on the original vocal and use floor to dial in the exact amount of attenuation you want. You might slide the key audio to the left about as many ms as your attack to get a smooth action. Optionally you can use the compressor rather than gate to attenuate the P's more proportionally.

This method takes a bit of time, but if there are enough plosives to justify doing it, it can give great accuracy - as other than the plosives, your key track will be all silence.

John

* Having contiguous audio for the key avoids gotchas related to Saw's keying if there are gaps.

Addendum:

I always use levelizer to remove silence - because it adds the attack and release times to the regions. It has been so long since I used the built in remove silence, - I forgot why I didn't use the built-in gate - and I started using this technique more recently.

The method I described should work fine with the built-in gate - I'll confirm this in the next couple of days.

One more thing that just occurred to me, you can automate the floor to dig into particularly loud plosives (or turn off erroneous activations) - there is no need for precision, just mark a random area around that plosive and pull down the floor a bit - the timing precision will already be in your key track - and the smoothness of the attack (combined with the left slide of the key) and release will nullify the abruptness of the built in gate. So it should be just as good - and even faster with the built-in.

When I realized that the storage for another 5 minute track of audio doesn't even cost a penny, I started using this technique for a variety of things (deessing, drum gating...). It is just a guess, but doing this with saw's built in processing is probably lower cpu wise than with a plugin.

One more note - use good track and file names to reflect what track your key is intended to be used for i.e. KeyLeadVocT17_L2Plosive. I also put them in a subdirectory like: TheSong/audio/SpecialProcessing.

[jmh]

After experimenting last night, I'm not sure I like the method I described for plosives. However I tried a modified version that worked pretty nice.

Engaged high-cut at about 80hz, and wrote that to a new file on a different track.
This was then dragged left & used as a key-source for levelizer compressor and or gate (reverse engaged).

I preferred this method for this task, because it was easier see the waveform to adjust the attack settings and left shift of the key track.

Notes:
the regular compressor & gate would have worked the same - I was just experimenting in levelizer.

My failure with the earlier described method may have been related to having my release too short and there may be something to do with the eq - I think I'll take another crack at it.
I have had very good results with that method for isolating drums.

[jmh]

Another surprise to me was that I was using much longer releases than I thought ~200ms - but now I'm thinking it maybe better with even longer releases.

I was talking with my hand in front of my mouth and you can really feel the air - and for a longer duration than I intuitively thought is would be. I never really considered the typical duration of the various syllables or the characteristics of how they evolve - or which ones tend to differ when sung vs. spoken.

Is anyone familiar with books about this?

[jmh]

Here is a link to a plosive experiment (I'll keep this for a month or two until I forget what it is for). It accomplishes the same as the method in post #8, but allows you to eq the attenuation (this is a directory with 4 .png screenshots and 3 very short .wav):

https://drive.proton.me/urls/J5B5DK64JM#IaULM1JgGcZr

When I recorded it (T1) I was hunched over (was set up for acoustic), and still used a popper stopper because I didn't want to spit on my new mic - held in my hand real close to the mic - and may have hit it. Anyway it is a bit nasty - and maybe it is ideal for this, as there is little consistency, as I was not intending to post this.

The procedure was as follows:

Duplicate track (T1) needing to be de-plosived, number selected the duplicate (T2), patch in eq high cut ~85hz (I used Nova - it has a higher cut off slope).
Build Mix to Current Hottrack, to create a helper-track (T4 - skipped one - I was going to try something else), slid that to the left maybe 20ms, and muted it.

Number selected T1, T2, and T4.
Inverted phase on T2.
Engaged the gate on T2, Keyed it to T4, played with gate parameters, changed Nova's high-cut & slope.
Engaged the Compressor on T1, Keyed it to T4, played with compressor parameters. (Notice that the compressor release is longer, and the attenuation is full band).



I wrote wav files of the original, and processed. To see what things looked like, I phase inverted one or the other, Number selected both, and wrote that to a Difference wav.
You can see some of the full-band waveform in the difference file from the compressor engaging (it had a longer release).

I did this pretty quick - if you don't count arranging the screen shots. Interestingly, I number selected the Unprocessed and the Difference tracks (this may need the phase inverted - depending on which track was inverted when created). Messing with the gain (and anything else) on the Difference, made for a simple readjustment - and may be an easy way to automate the depth of attenuation.

I don't think I'll use this method for a 10 second clip - but if I was doing a long dialog, I think it could dial in