I ditched Logic for Bitwig because Logic doesn't compensate properly for sidechain latency which makes any sort of precision impossible. It's turned out to be one of the best decisions I've ever made. With this level of precision, I'm finding I can really hone the details to get the exact result I want, and am currently trying all kinds of combos of multiband, m/s, etc.

I'll start by outlining some of what I've been doing along with notes of how it's going, and I'd love to hear from others what they've arrived at and why.

While there are many reasons one might want to duck, for the sake of simplicity, my examples will be about ducking bass with a kick.

1) MIDI triggering.
This is everything. It MUST be MIDI triggered. Even the fastest possible audio timed ducker still lets a few samples pass which causes combined peaks and defeats the purpose. I've run a bunch of tests, and in every test, MIDI triggering wins.

2) Ducking shape
Once I got used to MSEGs, I don't see why I'd ever want to use anything else. Drawing in the exact curve as opposed to a release knob with no real control is a world of difference. That being said, I've tested and shown that many of the commonly used ducking devices like Duck don't actually bring the level to zero when zero'd out. It attenuates, but not to zero. This means you can't ever fully eliminate combined signals. If you are using something alternate like an MB comp / dynamic EQ, have you worked with MSEGs? If so, and you still prefer some other method, why?

3) Frequency dependent
This is where I'm currently doing the most testing. Traditionally, people just slap it on the bass, but if ducking becomes a critical part of your sound, this may be insufficient for two reasons. The fist is that other elements may cause combined peaks, and the second is that ducking the higher frequencies in the bass may become unnecessarily noticeable.

The first solution is to do multiband, and only duck the lows. This works to an extent, and opens up the question as to whether this should then be applied to the whole music (and percussion and returns) buss(es) as opposed to just the bass.

The second is to do ducking on all bands, but with different settings, so the higher frequencies have a quicker release and the lows are suppressed longer.

I'd love to hear some informed opinions on how folks have this dialed in.

4) M/S ducking
I often find it useful to duck the mids only since the kick/snare tends to be dead center, and you notice the ducking more on the sides. This can help a lot, but doesn't always eliminate enough of the combined peaks by itself.

5) Hybrid (multiband and m/s) ducking
This is what I'm currently working on, so interested to hear from anyone who's doing something similar.

My current thought is that maybe the lows should be ducked entirely with higher frequencies ducked in mid only so side information passes. If you're doing something along these lines, how do you have it set up, and what lessons have you learned?

6) Group ducking
For any of the above, are you doing this per instrument like bass, etc or if you are doing frequency dependent ducking, have you achieved better results then applying to entire busses (instrument buss, etc)?

7) Ducking of (not by) vocals
Spectral ducking like making room for vocals is a whole other topic, so I'm asking here about ducking the vocals when something like the kick or snare hits... like a very quick duck just so the attack portion of the kick/snare isn't stacked with another waveform and a very short release. I haven't done this yet, but I plan on trying it.




You get the idea. If you're the kind of person with real-world experience doing this stuff daily, I'd love to hear your perspective and lessons learned about various techniques.

Interesting topic. Posting at this point more to follow than to have a great deal to contribute.

1) MIDI
I've gone in the other direction. Setting up separate 'ghost tracks' for kick and snare then having to remember to move them when I moved my real kick and snare got old fast. I much prefer setting up a s/c compressor (and, increasingly, TrackSpacer) because I can set and forget.
Re: double 'spikes' - since I started using clipping I've found that I can chop off anything extra I want downstream and stop compression and other threshold-dependent stuff further down the chain being triggered.

2) Envelope
My DAW's stock single band compressor (Steinberg) has a separate Hold and Release so I can sort of shape the ducking, but this is one of the reasons I got ShaperBox and I will be testing with this hopefully next few weeks.

3 & 5) Frequency dependent / multiband
Definitely an area I want to explore more. In my projects certain sounds (high stringy pads and leads so far) really don't take to even very delicate full frequency ducking for snare and other mid/high percussion, so I have set these aside to handle together on a test day. Will give TrackSpacer / Shaperbox / multiband compressor(s) a shoot-out.

6) Group ducking
For things like bass I did initially have each sound slaved to the sub via its own s/c and crossover point set in TrackSpacer, but the extra effort made them sound more distinct/separate. If you want a definite 'call/response' feel to a group of sounds, then this might be what you want, but usually I want a collection of sounds that are hitting the same frequencies to feel like they are all one very tweakable instrument, so treating them as a group saves time and sort of helps to glue them together in the way that a shared verb does. I also tend to leave TrackSpacer full range here for the same reason.

I set up pre-fader Send to a Bus with no output: one for every different thing that triggers ducking, then use Sends from those Buses to trigger the s/c. I then have all of these Buses in a folder, which I can then mute. My personal preference for ducking most of the time is that I want to notice when it's NOT on (rather than obvious envelope-shaping), so this allows me to check that I'm not overdoing it across the whole mix with one click.

I've not tried (4) but I really think I should

1) Run a test with something dropping to 0db triggered both ways and measure the peak that does (or doesn't) go through. I've never seen a case where MIDI doesn't win... unless maybe you want the length of time threshold is crossed to factor into the release behavior. I don't believe in dummy tracks either, but it's actually audio where you'd need a dummy track, then nudge it back in time to fix the latency issue. Study as much as you can how to work out MIDI sidechain routing in your DAW. It's time well spent. It's crazy flexible in Bitwig, I can route to or from anything on any kind of track, capture MIDI from multiple tracks into a single sidechain, etc. I don't miss dummy tracks.

2) Shaperbox is a huge step in the right direction, but the precision you'll be able to achieve is largely dependent on whether or not your DAW handles sidechain latency correctly. This is why I ditched Logic.

6) You mention Trackspacer a few times. I have nothing against it, but remember to ask yourelf exactly what you're looking to achieve with the ducking. If you're clipping signals, I'm assuming you're doing some sort of loudness workflow which means things like properly dropping a signal to zero before the first sample of combined peak occurs (MIDI triggering and MSEG) will be entirely more effective than something like Trackspacer which serves more of a spectral ducking / frequency de-masking purpose, but doesn't really help much with peak management.

Cubase's Frequency is fantastic for creative ducking, a dynamic EQ band can have external side chain input from anywhere - and multiple sidechains as well.

I also like playing with "ghost" (silent) side chain input for ducking, chucking random rhythmic parts in to create movement.

Would either of these work better for what you are trying to achieve?:

https://www.sweetsonics.com/plugins/laser

https://deviousmachines.com/product/duck/

I haven't checked out Laser, but it looks like it could if it does MIDI triggering. If I remember correctly, I think the issue I had with Shaperbox back in the day was the lack of MIDI triggering. That may be addressed now, though. Since moving to Bitwig, I'm using as many native devices as possible, so I haven't used some of these 3rd party plugs in a while. Bottom line for the most basic function is you need 3 things:

1) MIDI triggering
2) MSEGS (Shaperbox style bezier drawing tools)
3) Sample-accurate sidechain latency compensation

The third depends on your DAW, and the first two depend on the plugin. Duck hits both those criteria, BUT... when I tested it on Logic, I found it wasn't capable of actually attenuating to zero, so even if you drew the curve down to the bottom, it never fully avoided combined peaks. It attenuated quite a bit, but never to zero, and the difference was quite audible under certain circumstances.

That being said, I can see a use for audio sidechain ducking when used as something more than a simple trigger. If you can get a fluid and accurate read on the level and use that to auto-generate a ducking envelope (kind of what Trackspacer does on a frequency-dependent basis), then that opens up some interesting possibilities. I think this is possible in Bitwig's grid, for instance, though I haven't tried it.

I don’t remember when it was added, but Shaperbox (Volumeshaper) has MIDI triggering now.

Also in FilterShaperXL which can do interesting things in the space being discussed as well.

As an avid user of Shaperbox - what do you see FilterShaper doing that you can use Shaperbox for?

Technically the only capability not in Shaperbox is an LFO for Amount in addition to just the LFO for speed. You can use MIDI and sidechain with this LFO. Tradeoff is no keyswitch for the shapers :(

Also it's FilterShaperXL's encouragement for applying filters along with volume for ducking so you can have the volume come back up in ways that fit better. As the OP put it:

"The second is to do ducking on all bands, but with different settings, so the higher frequencies have a quicker release and the lows are suppressed longer."

I can use a Shaper to easily setup a high pass filter LFO to bring in frequencies in a controlled manner instead of just bringing the volume up. Yeah, you could do this in Shaper Box I just never did and never thought to.

I also seem to respond better to the GUI.

Subscribed. !!

I am assuming everyone on here makes EDM genre music.

Is there any technical reason why something like Trackspacer couldn't use a lookahead buffer to catch even the first samples of a peak? It would introduce some latency, but if only used during the mixing stage, it wouldn't be a problem, and there could be a ZL button to turn off the look-ahead for tracking. A limiter that lets peaks through before reacting wouldn't be particularly useful so it's a pretty standard feature on brickwall limiters, so why not in spectral ducking?

MIDI triggering requires a lot of extra work since the plugin isn't aware of the loudness contour of the signal, it's essentially just a gate signal. You'd have to manually draw the loudness envelope by ear, and if each hit is slightly different due to non-linearities (saturation plugins, analog emulation drum machines etc.) or humanization it's going to be different for every downbeat. I know producers who manually draw in loudness automation pixel by pixel without even using plugins, which is perhaps taking things a bit too far.

I'm actually making pop music, but with a decidedly EDM production style.

What I've found works for me is to us all the EDM production workflows to create the loudest music possible...

... then scale it way back to pop music loudness levels. Doing so results in music that is much punchier and hits much harder, yet is much cleaner and less distorted than the pop references at the same LUFS.

Really precise ducking is one of the key components of the secret sauce to do this. Most people just slap a ducker on and think its helping without ever actually looking at an oscilloscope. If they did, they'd see that most of the techniques people are using aren't avoiding the first few samples of the combined peak, don't have the optimal recovery curve, etc.