Massive instrument model - is messing with my mind.....

[Lindon]

So I have a problem, a client wishes to move from Kontakt to HISE and they want to transfer over their comprehensive kontakt engine into HISE, and its boggling my mind a bit:

The instrument can play up to 8 articulations all at the same time. Each articulation has a number of round robins in it, and up to 8 mic positions. These 8 articulations can be selected from up to 48 different articulations shipped with the instrument....

So all do-able so far - I just have 8 samplers each with an articulation loaded, with its RRs and its Mic Positions....

But examining their kontakt instrument I discover that each articulation has in effect two wav files for each note - and this allows them to cross fade between these two notes for "dynamics" either using the mod wheel or allowing velocity to do it for them.

OK I hear you say - no problem 16 samplers - in pairs to handle this different dynamics thing.

But.. they also want to do recorded legato on some of these articulations, so where am I putting
these legato sounds? Especially given I dont actually know which 8 articulations will be loaded at any given time. SO these legatos would need their own samplers too - lets say another 4 or 5 samplers....

All of a sudden I'm up to 21 samplers --- filled with fairly heavy weight processing - so I think this is a no go..... unless someone (anyone?) can offer me an alternative approach to this 2 wav files for dynamics thing at least. - I have to be able to cross fade the volumes of these two wav files somehow... any ideas?

[d.healey]

@Lindon said in Massive instrument model - is messing with my mind.....:

OK I hear you say - no problem 16 samplers - in pairs to handle this different dynamics thing.

What else are they using velocity for?

[Lindon]

@d-healey nothing. As far as I can see.

[d.healey]

@Lindon Good, that makes it much easier. You now have the entire velocity range to play with.

So you can map your articulations, and round robins across the velocities and use HISE's built in group xfade feature to manage all the crossfades.

If you run into issues with the large sample maps causing issues then we need to double down on nagging Christoph to fix it :)

[Lindon]

@d-healey said in Massive instrument model - is messing with my mind.....:

So you can map your articulations, and round robins across the velocities and use HISE's built in group xfade feature to manage all the crossfades.

I dont understand what you mean - I think I was sick that day group xfades was being taught....

[d.healey]

@Lindon You've been missing out. Check this video, the group xf part starts around 6 minutes I think but you might want to watch it from the start. https://youtu.be/0cn1l8231n4?si=FwVH2QOZA2lANNIT

The feature has improved a bit since this video but it's mostly the same.

[Lindon]

@d-healey yeah but see my first post

"The instrument can play up to 8 articulations all at the same time. Each articulation has a number of round robins in it"

So I need RRs and dynamics....and group xfades -- which I cant have with the xfade system

[Christoph Hart]

@Lindon Use RRGroup for the Xfades and encode RRs & dynamics into the velocity data slot. You probably don't have 127 velocity layers, so you can pack that information into one number.

So if you have eg. 4 dynamic layers and 8 RRs distribute them so that the first RR group goes from velocity layer 0 - 15 (0-3, 4-7, 8-11, 12-15) etc.

Then do the "decoding" in the script.

[Lindon]

@Christoph-Hart yeah...its ugly tho isnt it - and means there is a lot of messing with sample maps in a frankly unintuitive way - esp. for a customer who has no HISE experience....

I might try and convince them I can replicate and extend the idea of RRs with our script driven version (IRRE), but if not then its this unless theres another option.....colour me "unhappy"

[Christoph Hart]

@Lindon just reread the novel of a first post from you and you don‘t really have velocities right? Because then you can cram it all in 8 + x samplers for legato and call it a day.

You just need to convert the RR value to a velocity without multiplexing other information into the 7bit.

But no matter what you do I would recommend trying to leverage the xfade feature here.

[d.healey]

@Lindon said in Massive instrument model - is messing with my mind.....:

yeah...its ugly tho isnt it

I think it's elegant. It's actually an old technique from Kontakt. Greg from Orange Tree Samples wrote about it years ago and I've used it a number of times. https://www.orangetreesamples.com/blog/ksp-making-the-most-of-groups

[Lindon]

@d-healey said in Massive instrument model - is messing with my mind.....:

I think it's elegant. It's actually an old technique from Kontakt.

Old...., Kontakt...., those were desperate days - I was hoping we could move on....

[d.healey]

@Lindon Sometimes the old ways are good. What would you prefer?

[Lindon]

@d-healey either:

a way to address the volume of a playing wav file

or

xfade groups, that are a subgroup of RR or better still teh other way round....

I realise its a massive ask.

[d.healey]

@Lindon said in Massive instrument model - is messing with my mind.....:

a way to address the volume of a playing wav file

You can do that using Message.setGain() or Synth.addVolumeFade()

@Lindon said in Massive instrument model - is messing with my mind.....:

xfade groups, that are a subgroup of RR or better still teh other way round....

You might be able to effectively do this yourself by scripting the group management. I know this drop down for xfades only shows 8 groups

But if you click this button you'll see all groups have a xfade table:

It would be cumbersome to set all the curves via the UI, but pretty simple to do it via scripting.

I'd still explore the velocity route though, I think you're making it sound more complicated than it is. If you name the samples with the mapping in mind then the automapper will do all the work.

[Lindon]

@d-healey said in Massive instrument model - is messing with my mind.....:

@Lindon said in Massive instrument model - is messing with my mind.....:

a way to address the volume of a playing wav file

You can do that using Message.setGain() or Synth.addVolumeFade()

The problem here is these are fixed values, not relative and my previous experiments have resulted in volume jumps and dives depending on the incoming velocity....

@Lindon said in Massive instrument model - is messing with my mind.....:

xfade groups, that are a subgroup of RR or better still the
You might be able to effectively do this yourself by scripting the group management. I know this drop down for xfades only shows 8 groups

But if you click this button you'll see all groups have a xfade table:

It would be cumbersome to set all the curves via the UI, but pretty simple to do it via scripting.

Im really not at all clear on what you mean - I have two dynamic layers only, and up to 8 round robins per sound, how would I approach that?

I'd still explore the velocity route though, I think you're making it sound more complicated than it is. If you name the samples with the mapping in mind then the automapper will do all the work.

The samples are already named, at least for the first few instruments - they want them converting from Kontakt instruments..... so the sampling got done a long time ago, as did all the trimming, balancing and editing, and naming...

[aaronventure]

Here's an entire guitar with all RRs and articulations in one "group"

Other strings are in other groups.

This is incredibly simple because if a different guitar needs to be put in here, I just tweak the config script by defining the number of RRs, layers and the whole algorithm adjusts.

In your case, I would probably do the same: lay out everything in a few groups (because you have 48 articulations). I'd probably spread it all across 48 groups (makes it that much easier to manage) if the high-group-count-UX in HISE has been fixed.

When you have multiple dynamic layers, you would play both (by targeting them using your noteNum x vel scheme in the overall layout), and then use the newly added continuous per-event modulation to modulate the volume of each event, which allows you to define your crossfading curve using tables in ScriptNode. There's a snippet in the waiting room for it where you can see a UI slider modulating individual event's pitch. You can repurpose that to target the volume, define your tables in the modulator network and just select the different table based on the eventData value that you send with the event (which would, in this case, specify the layer number).

[d.healey]

@Lindon said in Massive instrument model - is messing with my mind.....:

The problem here is these are fixed values, not relative

If you use the current value too then it's relative: Message.setGain(Message.getGain() - 3);

@Lindon said in Massive instrument model - is messing with my mind.....:

Im really not at all clear on what you mean - I have two dynamic layers only, and up to 8 round robins per sound, how would I approach that?

There are many ways you could do it - I wouldn't recommend any and would recommend the velocity method, but...

You could put round robins for articulation 1, dynamic 1 into groups 1-8, then for dynamic 2 use groups 9-16. You setup the crossfade tables to that 1-8 fade out as 9-16 fade in - do this by scripting, not manually. Repeat for all the other articulations too, 16 groups for each. Then in your on note on callback you can choose which 16 groups are enabled based on the selected articulation.

@Lindon said in Massive instrument model - is messing with my mind.....:

The samples are already named,

Rename them if necessary, in the long run it might be better and batch renaming is pretty quick even for large sample sets. If I have a complicated renaming to do I get ChatGPT to spit out a bash script.

To match Aaron, here's one of my guitar maps, albeit quite a small one.

[Lindon]

@d-healey said in Massive instrument model - is messing with my mind.....:

You could put round robins for articulation 1, dynamic 1 into groups 1-8, then for dynamic 2 use groups 9-16. You setup the crossfade tables to that 1-8 fade out as 9-16 fade in - do this by scripting, not manually. Repeat for all the other articulations too, 16 groups for each. Then in your on note on callback you can choose which 16 groups are enabled based on the selected articulation.

OK I think I get it...

Lets use a real world example - I have a guitar articulation it has 5 round robins and cross fades between two set of 5 round robins. So I put set one in RR Group 1-5 and set two in RR groups 6-10.

I set up a CC Modulator (assigned to CC1 - the mod wheel) and using a table, the table starts high and ends low -- how do I attach this so it ONLY applies to RR groups 1-5???

[d.healey]

@Lindon said in Massive instrument model - is messing with my mind.....:

I set up a CC Modulator (assigned to CC1 - the mod wheel) and using a table, the table starts high and ends low -- how do I attach this so it ONLY applies to RR groups 1-5???

Once you enable group xf mode for the sampler you can click this button to see the tables for all groups. Each group gets its own table.

Then you add your CC modulator to the sampler's Group Fade chain.