I do something like this with the Hydrasynth. The 2.0 firmware lets you set modulation values per voice. I thought it was a bit gimmicky at first, but if you set it up to use different wavetables or scan rates per voice, you get some really interesting results.