The VSL tool is as fast as your midi interface and sound card...
Basically, real time. (A piano, btw, has anywhere from 20 to 50 ms
"latency" as well!)
I think the problem is that you're actually misunderstanding the way Synful works. It's not accurate to say "latency" -- it's not a matter of "how fast", but rather of how much information Synful is gathering about the performance. It uses a pre-defined 1-second "window" to analyze the musical material, and select the best sample (or resynthesis data in Synful's case) for the job. So it's not latency in the conventional sense.
The VSL approach is quite clever, but there are advantages to the Synful way, as well. The main one being when you have a series of short notes followed by a long, sustained note. I don't see any way for the Vienna Instrument to do this, since it's running in real time, save for using a new dimension in the matrix, and switching with mod wheel or some other controller. You'll notice, if you watch the video very closely, that the durations actually switch on the 2nd note of the new speed. The switching appears to be done by keeping track of the note durations, and using a speed threshold (the % slider) to determine the switch point. It's actually a very simple and elegant solution. But you'll notice that the last note is always the same articulation selection as the 2nd-to-last note. This is to be expected, because there's no new information coming in upon which to base a new choice of duration -- so the last duration used is selected. You'll also notice that the video only uses short-note articulations. I'm sure you could add a long-note articulation, but I doubt you'd get the long-note as the last note in a passage... I could be wrong, but I just don't see how it's possible, given the current state of space-time!!! [;)]
J.