Multi-Output vs. Single Output Plugins

In a recent thread I was enquiring about submixes vs groups.

Another thing that crops up is whether to run synths and samplers as multi-output or single output. I just thought I would canvas opinion on this too. Personally I have been leaning towards multi-output whenever the plugin supports it and I want to use more than one instrument. This means that Logic effects can be applied to each output. The alternative is to open a new instance of the plugin (e.g. Kontakt) for every instrument. It seems to me this would cause additional RAM usage - and one thing I have noticed is that it means it takes longer for your file to open as each plugin is loaded. It somehow seems neater to have just one instance with everything set up in one place too.

The only caveat with this is that you can run into issues whereby your volume on any track controls every MIDI channel on the plugin. A way around this seems to be to define custom controllers for something like Kontakt and send on those.

I'm just learning about this, not an expert yet, so any input appreciated, but that's what I've learned so far.

I've always wondered about this.

Personally, I tend to use a new instance of the plugin every time. However, a part of my brain thinks that on some level the Multi-Output method must somehow be better in terms of RAM usage...But then again....well how exactly? The only thing I can think of is that several instances of Single-Output potentially loads up more graphics (dubious theory).

The use of sound samples is going to be the same, the use of CPU is going to be the same, so what difference does does it make?

Drums (sampler) for instance, should be on individual tracks, but it is easier to edit midi on

one region than to have each pattern on diferent regions. I think it's more apparent on drums

than synths.

The only thing I can think of is that several instances of Single-Output potentially loads up more graphics (dubious theory).

I would think it loads up more than just graphics, probably an "engine" to some degree too. Instances in programming are usually not too "expensive" though, it is true, unless many, many are loaded. It's more important what each instance is doing. For example, if it has many samples loaded.

I would still think there is a bit of additional overhead though through loading multiple instances when one will do. As I say, I notice longer file load time.

In terms of performance once loaded though, I have heard the opposite. Namely that using multiple instances enables each to address a different CPU core, improving performance. Whether this is true I do not know, hence my post. I guess using multiple instances uses up more and more RAM (probably not too much) but might enable better CPU use.

I'm going to stick with the one plugin approach due to neatness of having everything in one place and faster file load time. I suppose this remains an open question.

Often-times I need to, say, side-chain an instrument's compressor to the beat of my kick drum, or wish to compress only a kick with a clap and none of the other compression. In these cases, I'll run Ultrabeat in multi-output and create Aux's for each track.

In regards to instruments like Reaktor, I tend only use Razor for creating specific one-shot effects or easily loopable instruments, and as such bounce to audio quickly. I rarely have a person use for its multi-output.

Another thing I have noticed about multi-output instruments is this:

They are often multi-input, too. So if you select "multi-timbral" when you create it in Logic, then Logic will create, say, 8 tracks with ascending MIDI numbers. You can then send MIDI to each part of the plugin, and set it to output on the 8 outputs you will also create using the plus button on the original track in the mixer.

In this set-up you can open the instrument window by clicking on the I/O of any of the tracks. You can record automation onto any of them too.

I have found that there is an alternative way you can do this. If you do not select multi-output when you create the instrument track in Logic but do selected multi-output when you initialise the instrument, the you go and click on the plus symbol and make the additional channels in the mixer. Then you right click each and choose "Create/Select Arrange track" and a track is added to the Arrange window.

Differences:

• The Core Audio for this is not the instrument but the aux channel.
  
• The instrument appears in the I/O but it isn't clickable - you must select it on the original track.
  
• You have to record any MIDI instrument-specific automation on the main instrument track. So, pan and volume can go on the aux track but other stuff on the instrument track.

I only just understand what's going on here and I assume it's better to go the standard route and record part specific MIDI data onto each appropriate track!