Kontakt - One big instance or a separate instance for each instrument?

I plan to use Kontakt for a dozen or so MIDI tracks in each song: bass, piano, drums (separate kick, snare, hat, etc.) and other stuff.

I trying to decide between using a single instance of Kontakt for all the instruments (e.g., Kontakt Library, Abbey Roads Drums, etc.) or a separate Kontakt instance for each instrument...

To me, the advantage of one Kontakt instance per instrument is that I would be able to more easily control the output buses and effects for each instrument.  Also, when I double-click on a track's soft synth icon to pop up that instance, I would see the correct instrument for that track, since there's only one instrument per instance.  If all instruments are within a single instance of Kontakt, I have to scroll up and down through Kontakt to find the instrument I want to tweak, which is a pain.

The disadvantage of one Kontact instance per instrument is the increased CPU and memory usage, correct?  Has anyone found out whether this is significant?  How much overhead is associated with each new instance of Kontakt?

Are there other pros and cons of using multiple instances vs a single instance?

Beagle


Kontakt's outputs are extremely configurable.  I would certainly (and this is what I do myself) recommend using 1 instance of kontakt and use kontakt's outputs and buses for any routing you need.

yes, using more than one instance of kontatk will be very taxing on your system.  I can't give you details because they're going to be specific to your machine, but every instance of kontakt will tax your CPU, your RAM and your hard drive access.

+1 to Beagle

I also thought about using multiple instances of multi-output VSTs for several reasons - one being that when you
freeze a synth, all outputs are rendered and frozen.  That was when I "needed to freeze" stuff all the time because
my system was just not powerful enough to cope with what I was asking of it.

The real solution to that issue was to get a more powerful system - done that now and I only freeze a synth when
I'm ready to, because I'm happy with the performance and want to work with the audio from there.

You haven't included your system specs, but since you're running Win7/64, my guess is you have a reasonably capable
system CPU wise.

Cheers - Cliff

I'm not too concerned about performance, since I just bought my PC a few months ago; it's a i7-950 w/ 12 GB RAM.  However, I've only tested about 5 instances or instruments so far, and would like to scale up to 12 or even 20 instruments in the future.

I'm a bit confused by Kontakt's output routing.  For instance, I'd like to try to apply a simple reverb to all my Abbey Roads drums tracks.  I eventually figured out how to do it via the Outputs panel, but selecting a reverb preset felt a bit clunky, and I found myself wishing I could get the drums to use one of Sonar's reverbs.  However, I don't know how to send just the drums to the Sonar reverb, since I only have a single instance of the Kontakt synth, and all its output goes to the same place.

Is the solution to create one of those "16 out" or "8 out" Kontakt instances, and then send each instrument to a different output from within Kontakt?  I'm guessing that would let me use Sonar effects on individual Kontakt instruments, which would take care of my first concern.

The other issue is how to quickly find the instrument associated with each track.  If I use a single instance of Kontakt, and double-click on my piano MIDI track, I just see the main Kontakt window, and have to scroll through the instruments in order to find the piano.  Is there a faster way to see the piano settings when I click on the piano MIDI track?  Or do I need to get used to scrolling within Kontakt in order to find instruments?

Thanks!

Cliff - I do it that way all the time!  If GPO won't do that then I am glad I never bought it - even tho I thought about it several times and ended up with Kontakt.

tripecac - yes, you'd need the 8 or 16 out of kontakt and then create the outputs in the kontakt GUI to correspond to the output tracks in sonar, then you'll also need to route the individual instruments to the separate outputs in the kontatk GUI by selecting the output on each of the instruments.

for things like reverb and compression, etc, you can select those on the outputs themselves (by using the pull down menus on the output mixer) or, what I do is not use kontakt for that anyway and just use reverb and compression in sonar using plugins.  I normally use a reverb bus and just send parts of each of the output tracks to the reverb bus.

as far as scrolling thru the instruments to get to the correct one, yeah, there's no way around that really. 

bitflipper


I was unaware that you could freeze multiple instruments independently. I always assumed you froze the whole synth.

I always route each instrument of a multi-timbral synth to its own audio track, be it Kontakt or Superior Drummer or SampleTank or TTS-1. But I have always observed that freezing means every voice on that synth.

So how exactly do you go about freezing individual voices?

Ah, I am very sorry, I meant to say "bounce."  freezing does indeed freeze all voices.  you can bounce individual ones, tho.

yes, you're right, bit.  color me embarrassed! 

Sorry, I don't have any screenshots or templates to share, but it's pretty straightforward. You have to first either load one of the multi-output (8 or 16 channel) versions of Kontakt or add some channels to the single-channel version, then specify which output to use in each voice's header. Normally you'd also assign a separate MIDI channel to each instrument.

As for drums, it depends on the sample library. Some have plain old GM mapping, and as such have limited routability unless you load multiple instances. The fancier ones have separate libraries for kick, snare, toms, cymbals, etc. so you actually load up the drums as separate instruments. They can then have their own outputs, their own MIDI channels, and their own effects, just like any other combination of instruments.

That said, I rarely route Kontakt instruments to separate outputs, because I just mix them in Kontakt and normally only need a single stereo output. Kontakt has pretty much everything you need in the way of effects, so there isn't a lot of benefit to routing to multiple audio tracks.

One scenario where you might want to do it, though, would be if you want to do a lot of automation for each instrument, in which case it's a little easier to do that at the track level rather than within Kontakt. Or if you have some special plugin you want to use on just some of the drums (e.g. a transient shaper, which IIRC Kontakt does not have). And I suppose if you were doing a surround mix then multiple outputs would makes sense.

I don't use Kontakt for drums, so I can't offer any suggestions in that regard.

Mostly I use Kontakt multis for orchestral ensembles and percussion, in which case each voice is a separate instrument, so each one's volume can be adjusted independently, either coarsely via the volume knob or more creatively via MIDI automation.

I don't know how you'd adjust each drum in a GM set - maybe you can't. Or maybe Kontakt has more sophisticated mapping features than I know about.

And yes, I have done exactly what you're describing: I've split out the hi-hat notes into a separate track and driven a separate instance of the synth with it. That also gives you the option of substituting a hi-hat from a totally different synth. But I've not done that for a Kontakt library, which seems like kind of a brute-force solution and poor use of RAM when you're talking about huge sample sets. But then, I don't use Kontakt-based drums, either, so I don't know any special tricks there.

The one to pose this question to is Jim Roseberry, who has actually created a Kontakt-based drum library and is therefore pretty knowledgeable about Kontakt's capabilities and limitations as pertains to drums.

Tripecac


Out of curiosity, what do you use for drums?

That is the question.
I use Stephen Slate drums in Kontakt and I can set the volume, tone etc of each drum within Kontakt. I can also swap individual drums with others as well as set each drum to it’s own channel.

Even for non-drum instruments, I am wondering how you handle multiple parts of the same instrument.

For example, if you have a piano sound, you might want to split right and left hand parts into separate tracks (or layers) for easier MIDI editing.  And ditto for a french horn section, or a violin section.  In each case, it's the same instrument, but different parts for it.

It seems you have a few options:

1) all parts on one track, separated via track layers
-- PRO: requires fewest tracks and least memory/cpu
-- CON: all parts share pan/volume controls
-- CON: editing individual layers isn't as easy as editing separate tracks

2) each part gets its own track, but all parts use same instrument (and MIDI channel) within the single kontakt instance
-- PRO: easier to edit than layers
-- CON: all parts share pan/volume controls

3) each part gets its own track and its own instrument/channel within the single kontakt instance
-- PRO: independent pan/volume controls
-- CON: increased memory/cpu usage
-- CON: need to coordinate patch or effect changes between parts
-- CON: need to coordinate midi channels between sonar and kontakt

4) each part gets its own track and its own kontakt instance
-- PRO: independent pan/volume controls
-- PRO: no need to scroll through instruments within kontakt instance
-- PRO: no need to coordinate midi channels between sonar and kontakt
-- CON: greatest memory/cpu usage
-- CON: need to coordinate patch or effect changes between parts

---------------

Are there other pros and cons that I am missing?

Which strategy do you end up using?

Do you have specific details on the implementation of Kontakt that quantifies how multiple instances taxes your CPU, RAM and hard drive more than a single instance...or are you surmising that this must be the case?

I did a test once with the task manager open.  Two instances of Kontakt definitely use more RAM (and I assume more CPU) than one.  Even using two copies of the same instrument within a single Kontakt instance requires more RAM.  It doesn't look like Kontakt re-uses the loaded samples within memory, but instead re-loads them with each new instrument.

I've tested it, watching memory usage as I added Kontakt instances and loaded libraries. Memory usage was about 40Mb per instance with no libraries loaded, which tells you the basic penalty for each additional instance.

Additionally, there is a significant economy when loading multiple libraries. After the first library, 80Mb of additional memory was allocated, but only 6Mb for each library after that. So it definitely makes much more efficient use of memory to load multiple libraries into a single instance of Kontakt.

In my test, if you loaded 16 libraries in one instance it would be a 120Mb load, but loading the same 16 libraries into 16 instances of Kontakt would take 1.9GB and would probably fail on a 32-bit system.

I can only speculate about CPU usage, as I did not test that. But my guess is there are some economies in a multi-instrument single instance. They may not be huge, but there is no doubt that a single instance would be more CPU-efficient.

I recall this being debated on the NI forums a while back. One theory doing the rounds was in a 32 bit system a single instance of Kontakt was desirable because of the memory limitations. multiple instances of Kontakt use more memory, so this bit at least makes sense.
 
However the theory continues that on a 64 bit system with more memory multi pule instances of Kontakt are better as there will be less CPU hit.
 
I am not endorsing the theory, just stating my recollection.
 
I had a quick look through the NI forum looking for the thread but did not find it. I did find a statement in one thread that said "Each additional empty instance adds between 40 and 60 MB to RAM usage." however.
As i ave not managed to stress my system yet I don't really worry about it. I will use several instances of Kontakt, each containing several instruments, it just seems to be easier for me to manage if you don't have too many outputs from each instance, so I tend load the 8 output version, unless there is a really good reason I want the 16 out. - But that's just me.

Windows does use both cooperative and preemptive multitasking, but the O/S has no control over task management for plugins that run within a host application. Windows also shares re-entrant code and only loads one instance for most DLLs, but plugins are managed by the host, not Windows. And for many plugins, the dynamically allocated memory far exceeds the memory footprint of the DLL itself, and the big-ticket items like data buffers typically cannot be shared across multiple instances.

Bottom line is it's not an OS issue. It's up to each plugin developer to make their products operate as efficiently as possible. Unfortunately, not all do.

You forgot one con to option 4: having to add two more video displays so you can show all your instruments at once.