Comment by tomcam

It's called modeling the human vocal tract. Sampling, then tweaking with DSP has proven to be a better approach.