I had to loopback cables once with my interface to record the output of Sonar for a video on how to use Sonar. It worked well.
 
Another member here pointed me to VoiceMeeter which can do the loopback virtually. I made a test video to try it out and it worked as well as the cables did.
 
Would VoiceMeeter also work for recording the live synth?
 
There are examples of how VoiceMeeter can be used on their website and it looks like it is one of the use cases. If it does it virtually, then I don't think it goes back through the interface and therefore would not add any (or maybe a little) latency?
 
Just a thought.

I'm pretty sure VoiceMeeter is Windows audio only, not ASIO. There are many non-ASIO solutions available for Windows.

I am completely confused now... More that 50 years ago, when "randomness"  was introduced in military/scientific/economical modelling on computers, they have realized that a possibility to produce exactly the same random sequence is a must, to debug/prove/resurrect all related calculations. The implementation is pretty simple: computer (pseudo)random generators have a seed, which is just a number. If you start generator with the same seed, it will produce "random" but still the same sequence. You change the seed - the sequence is different.
 
MIDI recording has no big sense without reproducibility feature in case there is any "external" influence on the sound (not only "random", can be simple wave over time). The solution is still the same, just "reset to X" signal is needed (X is a preset setting) to make MIDI reproducible and so still usable is such cases.
 
Yet I have not found any discussions about that topic, nor "reset all internal generators" option in Z3TA documentation... Have I overseen something or the whole music world has still not adopted that simple concept to solve annoying (as seen in this thread) problem?

I am no a synth-heavy person, but Z3TA+2 will get randomness simply from oscillator configuration alone (synched, multi, etc.). If the oscillators are not synched to each other, the sound will be different each time the same key is hit (depending where it is hit in the cycle).

fr: http://forum.cakewalk.com/record-a-synths-output-in-realtime-m3253176.aspx#3253176

That is what has convinced me to "change the lager".
Cakewalk has developed the DAW and the plug-in. Many Factory presets of this plug-in generate completely different sounds/volumes every time you freeze the synth. It is not about "subtle different sweet spot" or "different sound after moving the clip". It is about "why there is no sound on the place where it was, I have not changed anything!". That "feature" renders Sonar useless as a DAW for working with Z3Ta from the same house. I can not believe that was not noticed, much early that 2 years ago.
 
Before tests, I was almost sure that freezing the same untouched track will produce exactly the same result. Freezing is done with the whole track, always resetting all generators to some value on "time zero" sounds extremely logical for me and eliminates the problem.

Anderton
I'm pretty sure VoiceMeeter is Windows audio only, not ASIO. There are many non-ASIO solutions available for Windows.

It says it is ASIO unless I am misunderstanding what they mean by that

 

azslow3
I am completely confused now... More that 50 years ago, when "randomness"  was introduced in military/scientific/economical modelling on computers, they have realized that a possibility to produce exactly the same random sequence is a must, to debug/prove/resurrect all related calculations. The implementation is pretty simple: computer (pseudo)random generators have a seed, which is just a number. If you start generator with the same seed, it will produce "random" but still the same sequence. You change the seed - the sequence is different.

 
I think you are missing something - its not just the random seed that is involved. If a synth has a patch with an ever varying oscillator that drives one or more of its components (filter sweeps or arpeggiators) then the sound at any instant of time depends on when you play the note. On such patches since you can record the part on a track at different times from when the oscillator was started, it would sound different every time you played the key. Now when you play back the part the oscillator gets started at the beginning and sounds are generated based off the value of the oscillator at the time the note is triggered. However the start time of the oscillator is not guaranteed to be the same as when it was recorded so it will by definition sound different.
This is the most common scenario with Z3TA etc. Another is some analog synths are modelled to mimic electronic components using various techniques and such synths would never sound the same twice.

Noel Borthwick [Cakewalk]
If a synth has a patch with an ever varying oscillator that drives one or more of its components (filter sweeps or arpeggiators) then the sound at any instant of time depends on when you play the note. On such patches since you can record the part on a track at different times from when the oscillator was started, it would sound different every time you played the key. Now when you play back the part the oscillator gets started at the beginning and sounds are generated based off the value of the oscillator at the time the note is triggered. However the start time of the oscillator is not guaranteed to be the same as when it was recorded so it will by definition sound different.

And thus the request to record the live audio from soft synths so that it Does sound the same on each playback.

Anderton
I'm pretty sure VoiceMeeter is Windows audio only, not ASIO. There are many non-ASIO solutions available for Windows.

Voicemeeter (Banana) (how's that for a version name) actually does ASIO.