For example, taking dry VSL string samples and making it sound like live...
IMO, this is more a function of micro / macro dynamics. You can emulate an orchestra with samples, and if you're really, really picky with velocity and arrangement, you can get good results. However, samples simply can't achieve the inner-dynamics of trained musicians playing with each other in real time (reacting to each other and to the conductor). That's not even mentioning the minute details that vary every single moment that they are playing (as opposed to the same sonic signatures that have been captured in the files and are simply being recycled on a sample level). So that's one (and perhaps the most important) element of the "live" sound/feel that's missing from samplers.
The second is sound stage. If a sample library was captured in an actual orchestral hall, then the reflections of that hall are captured as well. However, reflections at one loudness are different than another, and as you arrange with different sections at different volumes and velocities (controlled via MIDI or DAW), you are hearing a misrepresentation of the original recorded audio, especially when more than one instrument are present.
As stated above, you can try moving instruments around on the sound stage separately, but you would not be able to position them in the original location that they were captured at. In essence, the early reflections from different surfaces of the hall would change if you physically moved the violin section - your samples have the same reflections, as captured by the mic, no matter where you place them on the sound stage. You can't consciously analyze these reflections, but your mind can decode them and that ability breaks the suspension of disbelief.
If it were a completely dry recording of one instrument at a time, you could place them at velocities and sound stage locations and apply the same convolution reverb to all, and that might achieve better results - but even that is not perfect.
So, you're kind of fighting against the tide. All this is not to say that you can't achieve a level of realism if you keep the above in mind. Adding yet another reverb (I say "another" because the original reverb of the original hall where the instruments were recorded is still present in the recording, remember) that's convolution-based can help glue things up a bit and mask these detrimental attributes.
It is also possible to add an enhanced stereo field of dry DI recordings and sample-based instruments with different mixing techniques, as well. For a nice exaggeration in the S channel(s) (S=side, as in everything outside of the very center of the sound stage), I really enjoy using the following in (very!) small amounts:
Console (or DAW without M/S plug-in) version (for all DAW users out there with M/S matrix encode/decode plug-ins, this process is much easier - I'll elaborate on that in a later post) :
- Send the stereo mix of your instrument(s) to a common buss and return to two grouped faders. Send the faders to the 2 Buss.
- Patch a stereo Aux (best if pre-fader) off the two-fader group we just created and return it on 2 more sets
of faders: M1 and M2, and S1 and S2.
- Keep both M's panned right up the middle (this is now your mono-summed Mid channel). You can group these, or assign them to a mono buss and return to a single fader for easier control (again, keep panned to center).
- Just as above, pan both S faders up the middle, but this time invert the polarity (phase) of one of the faders (typically the right channel, though it doesn't matter). This is going to be your Side channel, but it's not quite there yet. This is where it gets kinda tricky:
- Buss both
summed S faders to two mono busses, and return each buss on its own fader. Hard-pan these faders left and right, and again invert the polarity of the right fader. Group these together (don't buss just yet).
At this point, if you've bussed the Mid channel to a single fader, you'll now have 3 faders to control your M/S mix - the M channel will be completely mono, the two S faders will be your stereo field (these will completely disappear when monitoring in mono - but don't worry just yet). The more S you add, the wider your mix will be.
- Now, send all three M/S faders to a stereo buss (maintaining left and right for the two S faders) and return on two more grouped faders (panned hard left / right). Send this two-fader group to the 2 Buss (if your aux was pre-fader in step 2, then you can now monitor the L/R mix and the M/S mix separately).
- Patch another stereo aux on the M/S stereo fader group we just created to send to yet another set of faders. Group the new faders, but this time pan them opposite of the previous M/S group (right / left), and invert polarity on BOTH channels. Keep the faders all the way down on this group, because we're about to create a feedback loop...
- Set the output of this "inverted phase / pan" group back to the buss that's feeding the previous "non-inverted" M/S stereo group.
At this point, you should monitor your M/S mix ONLY at low levels and slowly feed the "inverted" M/S back through - you'll hear when the feedback is too much, because it will start to ring. Back it off a bit. You'll also notice your stereo image can be thrown extremely wide by using this technique. Remember also that any level changes before this point will affect the signal, so use caution when increasing levels in the rest of the chain.
All that's left now is to adjust to taste (don't forget the three M/S faders being fed by the original stereo mix*). Mix back in with the original signal, and you can get some pretty spacious results. A little goes a long way, so don't overdo it.
* = You can go even further with this by patching in a graphic EQ on the mono M and S channels created in steps 3-5. Whatever EQ additions or subtractions you make on the M channel, you should make the opposite changes on the S channel (in our scenario, you'd need 3 mono EQ's total). Best results if used sparringly and not past 600 Hz. This is a stereo-shuffling technique, sometimes used to make true mono recordings sound stereo.