Bouncing into stems

[Anders Peev]

I'm doing music for a commercial for swedens biggest milk producer and the film-people want me to deliver my mix in stems, naturally. Will my mix sound the same when the cinema-sound engineer puts my stems together if I just bounce them into the groups I think they belong into? Or must I be ware of something here? Will the limiter play me a trick when I bounce in stems?

Anders

[dix]

I always create a stereo mix and then A/B the stems (all set to unity) against the stereo mix to make sure there is no difference. The mixer can always refer to the stereo mix if he wants to hear what you intended.

Yes, I've seen people use elaborate side-chains to make sure the dynamics plugins behave the same on the stems as they do on the mix, but in my experience, as long as the dynamics aren't doing something radical, that simply bouncing/recording the stems is fine. You can always A/B the mix with the summed stems and check. …and subsequently tweak a dynamics plugin on a stem if the dif is too apparent.

[dix]

…also, I'd avoid limiting all together. And use compression lightly. Mixers (like mastering engineers) don't like it when you give them something with damaged waveforms that they can't work with dynamiclly. Again, sometimes it's stylistically appropriate to have the track heavily limited, but it will ultimately be up to the mixer how loud the track will be in the mix - so no need to limit to increase percieved-volume.

[Kubi]

Best thing to do is to set up the stems beforehand, especially if you think you're going to have to do something like this again.

I usually use four busses - DRbuss, INSTbuss, MELbuss and LYRbuss, for percussion, general instruments, melodies and layers (i.e. pads), resp.

Each gets a set of reverbs (usually one long one short, so a total of 8 reverb returns, two for each buss) and gets mixed in a sub master aux. So you end up with four main "masters", one for each stem.

During work those are bussed to the mains, but when you're ready to print those can then be bussed to their own stereo audio tracks directly, and voila.

Once you set something like this up to your liking, save as a template, and bam.



Ah, and agreed: Always also deliver a clearly marked full stereo mix as a reference.

[Kubi]

Btw, I now do all my work in stems, even if I'm doing a record and only end up prnting the stereo master. Just got too used to the process, makes for really easy mixing.

[dix]

How about dynamics Kubi? Do you ever side-chain compressors so that the stems all "move" together dynamically. I've seen it discussed, but never tried it. I'm not clear on how one would do that if they needed to.

[Kubi]

I don't do much full mix buss compression in general, but if and when I do, it's easy:

- Print an *uncompressed* stereo mix

- Put your favorite buss compressor on it, adjust to taste (or else maybe you have already done this on your Main Output's Master fader?)

- Now buss the (original, uncompressed) stereo mix to a new buss, let's call it "CompCtrl"

- Copy the compressor plug-in from the stereo mix (or from your master fader) with all settings intact to EACH stem, and change the control on each compressor instance from "input" to "side chain", and set each side chain to be controlled by your new buss "CompCtrl".

- DON'T FORGET TO DELETE THE COMP FROM THE STEREO MIX. Otherwise your driving the stem compressors with an already compressed signal. Or, if you set the mix buss compressor on the Main Output Master fader, delete it from there, so you don't listen to a twice compressed signal. In other words, your separate stem buss compressors, controlled by the (otherwise silent) uncompressed stereo mix, are doing the work now.

Voila, the stereo mix will drive each stem's compressor as if it were the whole mix, with full separation.

Btw, can't take credit for this trick, learned it from someone on this here forum or maybe on some other forum. But I think it was here.

[dix]

Yes! Makes perfect sense. Thanks for the tutorial!

FWIW, I usually prefer to just run the mix multiple times, soloing the appropriate tracks on each pass, to create stems. If I run it all at once I find that I want listen to the stems individually to QC anyway so it doesn't save all that much time doing them all at once. Plus I can't afford the cpu needed for multiple reverbs etc. However, the side-chain procedure you describe will work just as well for this one-at-a-time method.

Thanks again!

[leigh]

Each gets a set of reverbs (usually one long one short, so a total of 8 reverb returns, two for each buss) and gets mixed in a sub master aux. So you end up with four main "masters", one for each stem.

Kubi, why do you use short and long reverbs together?

**Leigh

[Anders Peev]

Thanks for all the input, and especially for saving me from putting a limiter on top of all! Thanks for that dix!!!

I found a very easy way to compare my stems to the intended mix: I put all my stems (4 of them) into a master chunk together with my reference mix. In solo mode I can toggle between the reference mix and the stems by klicking the play symbol of the reference track while holding in alt or apple. If there is any difference I would instantly hear it.

I will post the result as soon as its offical. It might take a while though, there are winter sport celebrities doing winter things in the film. So I believe it's not gonna be out until christmas.


Anders

[Kubi]

Each gets a set of reverbs (usually one long one short, so a total of 8 reverb returns, two for each buss) and gets mixed in a sub master aux. So you end up with four main "masters", one for each stem.

Kubi, why do you use short and long reverbs together?

**Leigh

Within a stem, some sounds usually need a larger space (i.e. a hall) some need a smaller space (i.e. a room.) I don't usually use them together on the same sound, but I need a minimum of one each within each stem.

I found a very easy way to compare my stems to the intended mix: I put all my stems (4 of them) into a master chunk together with my reference mix. In solo mode I can toggle between the reference mix and the stems by klicking the play symbol of the reference track while holding in alt or apple. If there is any difference I would instantly hear it.

Here's another quick test:

• Put the full mix and the stems in the same sequence at the same (unity) level.

• Instantiate a trim plug on the full mix and reverse polarity.

• Press play with all open.

You should hear complete silence, or more likely, just the remnants of odd reverb tails with zero dry sounds (since often times reverbs are non-linear and therefore do not print the exact same way twice.)

If you hear anything more than that, the stems and full mix are not identical (i.e. an instrument you accidentally forgot to put in a stem would show up clear as day.)

[dix]

Revisiting this thread, because I'd like to deliver stems as described - with the dynamics of the entire mix imposed on each stem so they sound identical when compared to the mix.

Kubi, you say

- Copy the compressor plug-in from the stereo mix (or from your master fader) with all settings intact to EACH stem, and change the control on each compressor instance from "input" to "side chain", and set each side chain to be controlled by your new buss "CompCtrl".

But only certain plugins have side-chain inputs, correct? I thought placing the Dynamics plugin, which does accept side-chain info, at the top insert might allow me to have the other plugins in the chain, but that doesn't work (seems like it should ). Is there a way to have any plugin receive side-chain input?

The above stem-phase-check is genius btw! Patch randomization from unfrozen VIs can throw it off, but this is a very useful trick.

[outofhand]

You should hear complete silence, or more likely, just the remnants of odd reverb tails with zero dry sounds (since often times reverbs are non-linear and therefore do not print the exact same way twice.)

Kubi, I just noticed this comment - can I ask you about it a year after you posted?

I've spent the last several months revisiting some projects that are 10 years old.
I'm re-bouncing these projects on the same computer and software used when they were first done, with the exact same settings.

But I've been noticing that the reverb tails don't null with the originals when I reverse polarity and compare, and haven't been able to figure out why.
Is it true that algorithmic reverbs can contain internal modulation that is hidden from the user, which results in slightly different results with each pass? ie. not printing the same way twice, as you mention.

I was using the Waves Trueverb and Renaissance Reverb.

Wondering if you know more about this, or can point me at further reading. I've been finding it very difficult to track down.

cheers