Features we wish Reaktor had.

Snapshots and Presets are different things.

There are instrument level and Ensemble level snapshots, sometimes instruments are embedded.

How can you differentiate with just a simple 'click a button to save snapshot'?

Presets already do that. A preset loads the ensemble it was created with, You can organise your presets however you like. You can create folders base don categories. Just use your OS file system. Very easy.

I wish there were a convolution module in primary, where you could load files to simulate rooms, cabinets, and all kinds of other things. The lack of convolution in Reaktor has always been a head-scratcher for me. It's an amazing synthesis tool, and NI's clearly good at doing it. But for some reason it never came.

I want iteration and voice handling inside of core.

Actually, I think if Reaktor can add Sine Bank module and Modal Bank module, why doesn't it add a module about FFT?

Iteration in core is the biggest long term missing feature. It would be the basis for various other powerful missing mechanisms, particularly relating to code reuse.

Limited voice handling is already possible. You just import the voice ids from primary, and roll your own.

Voice manipulation isn't possible. You can't swap voices, or access values from one in the process of another, but I think that's impractical given the Primary -> Core hierarchy. Core can't really change any Primary level constructs. I think allowing that would be too costly in cpu terms.

It's optimised for passing events and audio events between the two layers, and even that bears enough of a cost to make it something we try to minimise.

What would you do with FFT?

Exactly!

Fft on its own isn't particularly useful. It's a building block, and applying it successfully almost always requires some advanced DSP.

The complexity of suporting code needed makes core a much better fit in terms of maintainability, and cpu efficiency.

It would make much more sense for Primary level to have specific fft based processes like convolution, or phase vocoding (even that is more of a building block though). But we know that new stand alone Primary modules were not a thing even years ago when other updates were still happening. The new factory components in the last big update - oscillaters, filters, granular etc. - were all built as core cells. There were no monolithic high level Primary modules. Primary has been legacy for a very long time.