If the demos sound like the LSO , then i'll save my money to buy this ...
It seems their sampling approach , for let's say the violins, was to record desk by desk ( one desk is 2 players )
Then , apparently, you can achieve a chamber strings sound or big strings sound , what you want... but i'm septic about this approach, Garritan said on another topic that it worked well with Brass , but what about strings ?
I remember many people here tried to get a section sound by using several solo violins samples together, to me the result was so-so , i didn't like the vibrato which's often too much pronounced when using this method.
I too am skeptical of this approach. I don't see how mixing 8 seperat desks together can sound like 16 players all playing at the same time. You are losing the natural harmonics that occur when mulitple players play simul. I guess the demos will shed some light on this.
From the demo it looks like Garritan's ensemble builder approach with a few extra smarts and reverb built into a single UI. I'm curious to know what the system requirements are - and of course, how much they're going to want for it. When they're talking about how this system will help them replace $30,000 worth of copyists - it makes me wonder what kind of perceived value they're trying to sell.
Personally, after the reports from the CES show on the new sample player from Garritan (and Plogue) I'm more interested in seeing that in action. It seems like they've got a UI cooking with drag-and-drop of instruments to a virtual sound stage, perhaps laying the groundwork for Garritan's Real Spaces impulse library. That to me sounds more realistic (from a resource and realism perspective) than AI's demo where they describe having everything and the kitchen sink loaded just to build a small ensemble. It'll be interesting to see both at NAMM this year.
As far as the "harmonic" effect of individual instruments in the same space is concerned, I think the real-world effect of this on the overall sound is over-dramatized in these forums. The AI approach seems to do a good job of masking a lot of other stuff that makes virtual performances sound robotic, such as ensemble lag and variation of vibrato amongst the virtual players to get a more effective sound of "smearing" that happens in real life. Of course, this was already effectively demonstrated by Francesco Marchetti with GPO but it looks like Audio Impressions has made a lot of the work happen automatically, which is nice. I don't think they're interested in replacing live players as much as they're interested in being *that* much closer - and replacing expensive copyists.