I'm going to answer this as if "channel" is reimagined as "Part". No, scenes cannot change the AWM2 samples used within a single Part. The closest you could get is to have a scene move superknob and have that action raise/lower the volumes of elements within a single a Part to get a different mix of samples (or operators in FM). This doesn't save polyphony however.

Currently, it takes non-zero time to "load" a Part with a different sound (parameters, samples, etc). You can do this manually by using MSB/LSB/PC. I take it any implementation of this feature would be as slow - so, as a thought experiment, if the feature were implemented you'd have to "wait" some amount of time after pressing the scene button before the new sound is available. Not a huge amount of time - but it's not instantaneous.

Recently, I was working with a set of four Element AWM2 sounds... (both the original Motif and Motif ES were 4-Element AWM2 synthesizers). Loading the Motif ES sounds was, firstly, an eye-opening experience. Sounds I remembered from 19 years ago seemed to jump out of the MODX/MONTAGE engine.

Since every sound was 4 Elements or less, with very little Knob and Controller assignments (which were limited back then) I started building PARTS that combined two 4-Element sounds -- which I proceeded to program into all manner of splits, layers, split/layers, morphing sounds, etc., etc. thus allowing two instrument sounds to exist within a single Part.
Definitely a rabbit hole I spend much time exploring - worth the exploration (imho). By placing two instruments in the same Part, quite naturally, you must workout how they are to interact - because they must share some parameters. Each of the Element groups can have one of the Insertion Effects (which is interesting). Because you can route each Element to one or the other of the Insertion Effects running in parallel.

There are examples of Single Part Performances in the MODX/MONTAGE that include several instruments... You can design which Controllers are going to bring in which Elements - that is totally up to you.

For example, the String Performance "St Ensemble K"
This Single Part Performance features a 5-Element setup, Elements 1, 2 and 3 are string ensemble sounds, Element 4 is Brass Section (hidden on MW), and there is a little Bell sound on the upper keys when you engage (AsSw1)

You can select any controllers you desire; if you want use the SCENE function to recall what is sounding, then you want to work with SuperKnob/AssignKnob system to work the Element Levels. I've never really thought to use the SCENEs to switch between sounds, but it is certainly within the functionality of the system.

I was thinking (control) arpeggios could be used but was slow to come up with this when I wrote before. [ASSIGN 1&2] (the two assignable switches) mentioned by BM dislodged the idea.

Scenes can change arpeggios and control arpeggios can turn on/off the assignable switches. So using this you can turn on/off elements a better way than using levels. This would preserve polyphony since some elements will not consume polyphony when XA control is telling an element it should be "off" due to the current assignable switch setting. There are built-in control arpeggios that set the assignable switches (IA1_FolkRock AF1 _N, IA1_FolkRk AF1&2 _N, MA_Dream Trance AF1, MB_Dream Trance AF2, and more ... search for AF1 or AF2). AF1 and AF2, fyi, are how assignable switches were labeled in previous generation keyboards so this is carry-over of shorthand. New docs abbreviate A.SW1 and A.SW2 - but that still is more verbose.

"IA1_FolkRock AF1 _N" and others not in the Ct/Hb category are hybrid -- notes and control. The benefit of these hybrid ARPs is that they don't do extraneous controller movements like many of the Ct/Hb do (more than just A.SW1/A.SW2) and the note part of the arpeggio will be filtered when you change the arpeggio mode to "direct" (so no notes are "played" by the ARP like when using "sort" ).

So setting the arpeggio to hold=ON, individual to "IA1_FolkRock AF1 _N", sort mode to "Direct" - when the scene is using this arpeggio, A.SW1 will turn on and when not using this arpeggio (when arpeggio is off or not running) then the A.SW1 will be off. Assuming A.SW1 is initialized to OFF to begin with. And it doesn't matter if the A.SW1 (aka [ASSIGN 1] ) is set to momentary or latch because the ARP control of A.SW1 will work the same either way.

I think this method is superior to using superknob and levels. This also frees up superknob for other things while saving polyphony.

You have options.

thanks for quick answering my questions.
I think Bad Mister' s Idea came closer to my wishes. With 8 Elements each Part I could split different sound by region or maybe better solution by velocity so 1 until 64 is el Piano for example and 65 until 127 is ac. Guitar. Then I must change the velocity of arpeggio to to play the sound I want to.

I will give it a try

Thanks

Regards

Yes, if you want arpeggios to play different sets of elements in a single Part then velocity map the elements and leave them on all the time. You can even use scenes to change the same arpeggio (if you want) to another of the same arpeggio and set the velocity offset on each to +/- percentage to target one set of elements vs a different set.

Or roll your own arpeggios with specific velocities "built in" and call up these specific velocity arpeggios (with no velocity offsets) to target the elements you want. You possibly reduce your dynamic velocity range - but that's a tradeoff you can manage.

I'm tying back to your scene request.