The time constants of any sidechain-controlled effect or utility will depend very much on the nature of the effect.
For instance, the attack time on a noise gate (which turns things "on") will generally need to be a bit slower than the attack time on a limiter/compressor (which tries to rein in the peaks). The attack and release time of envelope-control of phaser speed will need to be slower than the respective time constants of an autowah. And so on. In some instances, having the desired change occur in a very punctate way can be jarring and disruptive, while in other cases if it is too slow it won't accomplish what needs doing.
So, envelope-control over morphing into, and back from a loop, will require some thinking and probably trial & error testing, of the time constants of the sidechain. It obviously can't be too responsive or else "mis-picks" will bring the loop in at times when you don't want it. It can't be UNresponsive, either. The right degree of responsiveness will also rest on how perceptibly smooth the transition is. I get the sense from the OP that what is desired is not a complete
changeover, but rather a proportional blend. That is, the harder I pick, the more apparent the effect is. The "clean" signal may well never completely disappear, but the loop processing just becomes more obvious.
My sense is that one needs a larger averaging cap for that, which will effectively "sum" successive strums, and not reach maximum with a single pick/strum...at least not often...and gradually fade back.
At first blush, one need only use an LDR in parallel with a mixing resistor in an inverting op-amp/mixer configuration. Let's say Rf = 100k, R1=47k, and R2=470k, with R1 feeding the non-loop signal, and R2 feeding the loop. The gain of the non-loop is a little over 2x, while the gain of the loop is much less than unity. If our envelope follower drives a vactrol, whose LDR is in parallel with R2, picking harder (and of course using a suitable LDR value) will reduce the combined parallel resistance of R2 and LDR, making the loop more audible. Get the time constants right and you'll get a pleasing fade-in/fade-out. But here's the thing: it will also increase the amplitude of the combined signals. And that may be disruptive. The objective, then is to change the mix
without changing the volume level....unless that's what you want.
