Josh Steiner wrote:
> 
> Hi all,
> 
> Could someone tell me how you filter to get that legendary "talk-box"
> sound? (think: Intergalactic Planetary by the Beastie Boys) Perhaps how I
> can do it live with Csound and how i can do it with normal filtering
> programs (say, cooledit).

Beastie Boys probably used an analog vocoder.  You could write an
algorithm in Csound to generate this effect in a similar manner, but it
would be huge - 14 to 22 input bandpass filters, each followed by an
envelope follower, and the same number of output bandpass filters, where
the output of each filter is multiplied by the envelope follower
corresponding to the proper input filter.

A better way to get this sound in Csound is linear predictive coding. 
Use LPANAL to analyse a vocal track (preferably one that has no reverb
and is as clear as possible), and use the resulting analysis file in an
orchestra with lpread and lpreson to filter the source of your choice. 
For the source, buzz or gbuzz works well for pitched sounds, and rand
produces a nice evil whispering sound. 

As far as LPANAL goes, I have found that specifying larger analysis
window sizes works better. Specifying 500 (which results in a window
1000 samples wide) for a vocal sound recorded at 44.1 khz works well. 
Larger numbers of poles results in a clearer speech sound, but may not
sound properly "robotic" for your needs.

You could probably use an FFT based technique to get the same results
(phase vocoder, some of the other FFT opcodes I haven't tried), but
linear predictive coding is great for that vocoder sound.