Well, the thing there is that you'd have to re-synthesize the voices, wave-shaping wouldn't be enough, so what you'd basically want is a vocoder. The problem,

Well, the thing there is that you'd have to re-synthesize the voices, wave-shaping wouldn't be enough, so what you'd basically want is a vocoder. The problem, of course, is that synthesis of human voices isn't advanced enough to handle those kind of dynamics, so the problem where you'd end up focusing is on voice synthesis.