MIDI-to-Speech Converter?

[Kyle Jarger]

aschrock at cs.brandeis.edu wrote: 

> This probably isn't *exactly* what you're thinking of, but other than the
> midi control you're describing a vocoder exactly. (what kraftwerk used)

I believe that what Dr. Forreseter is referring to is NOT a vocoder, 
but really a singing machine.

They are machines based on early speech synthesizers, 
which used phonemes (the basic units of sound)
, and sound very synthetic, kind of like from the 
ad for that movie "War Games" a while back----
" Do you want to play a      Game?"
I think Texas Instruments made the ICs.

These communicate via a serial port, and you can actually program the 
pitch and timing of the words, too.  Kraftwerk's was a step up, as it 
can do glides, too,  Check out "Crosstalk" on Karl Bartos'  Electrik 
Music "Esperanto" CD for an excellent example.

Komputer said that a friend built a custom one for them,
 they painstakingly program theirs for their songs. It's solo'd (and 
pretty freakin' cool if you ask me :  )   ) in the 
beginning of one of the remixes of "Valentina" on that CD.  

I have one of these boards (from a ham radio flea market)
, and the speech IC needs about a 40KHz clock to run.

The board has a very high frequency clock divided down like a 
top-octave divider synthesizer so that it can actually sing tunes.
So theoretically, it can be used to sing tunes in a limited way by 
programming a file and just sending the data to it (You an program 
rest times, and sound phonemes multiple times to "stretch" words.  This sounds like 
what Komputer does. It couldn't possibly stay in sync with a track 
for very long using this technique.

To do it right, sync wise, you'd need to have the phonemes sent serailly from a host 
computer at just the right time, so perhaps on a computer, you could
write a program which recognizes MIDI clocks, and in response, acts
like a "serial data sequencer", to send out the ASCII character
strings to the speech chip at just the right time.

Perhaps I could make it "sing" along with a pitch (w/glide, bend, 
vibrato) if I could 
reliably produce a pitch about 10 times higher than normally produced 
for musical purposes, and feed it into the speech chip clock signal.  
I distinctly recall that the serial port and speech chip had separate
clocks, so this might be possible...
Perhaps an 8-10X multiplier PLL can do this well enough from a 
regular audible frequency wave ouptut from an analog or digital synth?

Regards,

Kyle Jarger
jkjelec at tiac.net
JKJ Electronics
Visit our MIDI/CV website http://www.tiac.net/users/jkjelec