Lars Luthman wrote:
> 
> I was listening to my brother's old records yesterday and found a sound
> that I really liked: The words between the verse and the "computerwelt"
> chorus in Kraftwerk's "Computerwelt". I'd love to have something like that
> in my music. Does anyone know how to make "normal" voice samples sound like
> that? I suppose it can be done with some kind of vocoding, but I don't know
> much about such things.

As other people have pointed out, the most accurate way of producing this sound
would be through one of the speech synthesis programs you can download, or
perhaps using an actual Speak 'N' Spell. Try using a number of different speech
programs, as this would be a good way of generating the different characters
heard on songs such as "Numbers." 

If you wish to modify vocal samples, the most accurate way of reproducing this
sound would be linear predictive coding. Do an LPC analysis of the speech
samples, then use the analysis to filter a buzz or gbuzz waveform. LPC can sound
very robotic, which is often a detriment, but would be a good thing in emulating
the voices on this Kraftwerk album. Be sure to start off with clean, dry vocal
samples, without any reverb, spoken slowly and clearly (i.e. talk like a robot),
in order to get the best analysis.