[LAU] Voice Synthesis

[Gordon JC Pearce]

On Sat, 2010-05-01 at 22:35 +0300, Chip VanDan wrote:
> Hey there again,
> 
> I was wondering if anyone out there knew of a voice synthesis program
> that could be routed through Jack?  Maybe I'm not using the right
> terminology, but I want to be able to make the computer say what I
> type, but have the output routed through different effects processors
> via Jack.  I'm losing sleep over this, but I have this choir of robots
> in my head that need to get out.  All the ones I've found so far work
> completely independently and recording them is no less than a superb
> pain in the rear end.  Perhaps it's time I learn how to program so I
> can make my own singing robots.
> 

The way I did it was I dumped the output of festival into wav files, and
then loaded them in specimen.  You can feed festival an XML file that
describes the pitch, duration and voicing of a syllable.  Example:

http://www.gjcp.net/~gordonjcp/mp3s/robots.mp3

As Julien said, the other option would be to use a vocoder, which
applies a filter shaped like the frequency response of the modulation to
a carrier signal.  There is a LADSPA floating about for that, but it's a
bit buggy.  I modified it:

http://www.nekosynth.co.uk/browser/wip/vocoder/trunk

It really needs the lower band replaced with a lowpass filter and the
highest band replaced with a highpass filter.  It sounds like this
(carrier produced by nekostring):

http://www.gjcp.net/~gordonjcp/mp3s/kelly.mp3

Ideally you want a bright buzzy carrier with lots of harmonics to
filter.  Like, uh, a string ensemble...

Gordon MM0YEQ