I'm currently trying to put some speech through a vocoder and add a melody. I'm using Reaktor 6 and the vocoder from this collection. I connected an oscillator (and sequencer) to the carrier input and the voice to the modulator and it works fine, except it isn't clear. The speech isn't 100% intelligible and especially the m's and the n's get completely eaten. I noticed it helped if I use a sawtooth wave and add more oscillators (there's 8 now), but it isn't good enough as the m's and n's are still gone.
Are there some rules on how to do it?