From Gabriel Heinrich:
“The Virtual Sound Stage 2 design assumes that the input signal is stereo, and relatively decorrelated already, which is true for most popular sample libraries.
If your input signal is mono (Sample Modeling Brass, for example), or the left and right channel are highly correlated (this is the case with Sample Modeling woodwinds and strings, and occasionally very dry sample libraries) you will probably want to add some processing before VSS. Adding a short IR with a high wet/dry ratio and absolutely no pre-delay will usually do the trick.”
Gabriel suggests introducing some decorrelation into mono signals prior to VSS2. This is due to the fact that phasing and comb filtering are much more problematic when we delay one of two absolutely identical channels, which is exactly what happens when SM Brass instruments are routed into the plugin. The short IR will help decorrelate the mono signal, making it ever-so-slightly stereo, and will decrease the appearance of phasing and comb filtering in VSS2 by adding variation into the left and right channels.
If you happen to record live performers in mono, and very dry, you may want to treat them similarly before routing to VSS2.
However, the Sample Modeling SWAM and SWAM S instruments (the woodwinds and solo strings) are a bit different. They all include various “instruments,” which are actually impulse responses that model different microphone setups and positions. This means the SWAM instruments are highly correlated, but not 100% mono, and therefore you may not need to add a short IR prior to VSS2 to get them sounding appropriate. They already have IRs built in.
It should be pointed out that these tips apply to virtual instruments other than those by Sample Modeling, and your own live-recorded soloists, or sections. I’m using Sample Modeling instruments for my examples here because I use them heavily in my own work.
If you have any questions about this topic, or any related topics, feel free to comment or email me at email@example.com.