This is a really interesting paper on VOSIM speech synthesis that I think was posted to the old forum.
Definitely a concept worth revisiting.
Thought I would also include a link to the FOF method, since the two seems to pop up in the same discussions all the time:
Crazy example of vocal synthesis from IRCAM circa 1983.
Is this related to chip speech?
I was playing around with their free vst a while ago. They sell personalities.
The free one I was just enjoying — Alter/Ego isn’t so much a throwback, but it is about singing.
Pulse Train Generator.audulus (8.9 KB)
Here is my first pass at making the fundamental building block for VOSIM, which has an number of sin squared pulseforms (N) of a given length (T) followed by a period of silence (M). The one part that needs some work is the rate at which the pulses diminish in amplitude (B). Ideally each pulse would diminish by a percent of the amplitude of the previous pulse, but this prototype uses a linear function and B controls what percent is mixed in.
I have also started messing around with an oscillator version, since working in milliseconds is probably nobody’s favorite. But I have a long ways to go on that. Until then here is a patch that’s just messing around two detuned VOSIM pulse generators and some drums.
Playing with Pulses.audulus (935.1 KB)
LOL, there’s a whole tutorial on how to build a vosim oscillator in Reaktor. I might stay up and try and build along in Audulus.
Edit: Well I made it but I gotta say I’m not so impressed. This version is fixed at two pulses and has some nonsensical scaling that causes clipping. Still I’ll include an example for completeness sake.
Well I politely inquired how Plaits voice synthesis (Why I’m interested in Vosim to begin with) worked and I got a really complete reply!
