Is it possible in Dorico to route a voice to two VSTS simultaneously. In particular I’d like to send each vocal part to Note performer and Cantai and mix the result.
If it is possible, how do I set it up? (I’m not very technical, sorry)
Ideally I want a template that applied that automatically.
[I’m thinking this might be a useful way of mocking up a passable choral sound without having to wait a very long time for Cantai to itself process its own multi-voice (?) choral voices]
No, it isn’t possible. The simplest workaround is to just have two staves with identical notation routed to the two VSTs and you can mix the results in the mixer - obviously excluding the duplicate in the layout you’ll print from. Other alternatives are 3rd-party VST hosts like Bidule or Vienna Ensemble Pro that have the two VSTs listening on the same MIDI port and channel.
Perhaps others have better ideas…
I’m not sure how it would work with those two in particular- you had to mention the ones which probably have the most challenges 
for other reasons. You can though use one of a number of tools which can split and route MIDI. The basic set up is that you send the midi from the Dorico instrument to the router first, whether it’s one that is a VSTi plugin or to an external port/program, and the router sends it onto both destinations. Works for lot of things others than those two - but I think those wouldn’t work out.
That’s a pity. There is a huge difference in sound between a choir and soloists especially when they combine in something contrapuntal. Choir lines basically have much more “basic tone fundamentals” in them, so today the combined vocal lines from Note Performer can give quite a good impression of how counterpoint will sound; although obviously lacking the consonant attacks and differing vowel syllables.
Today the AI approach to generating a choir sound seems to be based entirely on the idea that you can apply your AI to say six voices And obviously you get a decent choir sound. Unfortunately most of us don’t have the kind of hardware that can calculate that in real time. Which is why I wanted try the blended pragmatic approach.
Massive hardware for doing AI does exist of course - I used to work commercially for a company which had a grid of computers - but while this is OK for film companies or massive corporations it’s not something the rest of us can afford…
There comes a time, long past depreciation of the asset (powerful server), that the (my) company no longer needs it or has replaced it for a newer “race car”. That’s where I step in. They use a noticeable bit of electricity, have slower CPUs (usually) but in my latest grab, there’s 64-cores, 256GB RAM, 10s of terabytes of storage in a powerful RAID configuration.
… and it doesn’t end up in a landfill.