Physical modeling using audio input in Tassman / Prism

[KTlin]

Hi guys

What I would like to do is somehow recreate the synthesis in Madrona Labs's Kaivo, using Tassman or Prism (or other physical modeling synth). Kaivo uses samples in a granulator as exciters. What i want is to use the audio from a track, in real time, as the exciter. I know Tassman and Prism can both be used as effects in order to process audio, and I tried using them like this, but I got unpleasant results.

If some of you managed to get interesting sounds using this technique, I would appreciate some advice.

Cheers