xoxos has released fauna, a VSTi which emulates animal vocalisations by inducing oscillation as a function of structure instead of processing a source oscillator.
fauna is an elementary application of the kelly-lochbaum tract in conjunction with a simple reed model.
The kelly-lochbaum vocal tract employs a series of coefficients representing open area of the vocal tract to compute transmission.
The length of the tract is divided into five sections, or waveguides. these waveguides function as conventional delays set to very short times to model the acoustic transmission through air.
xoxos explains fauna with a pretty illustration.
The low waveguide count allows a variety of animal sounds to be mimiced to some degree at a low cpu expenditure.
The glottis functions as a valve, opening with pressure from the lungs and closing with pressure returning from the mouth. The smaller the aperture, the higher the reflection.
The throat of the tract is tapped separately and passed through a resonant low pass filter to add throat or body resonances. These are key components of some tones. The signal at the throat is generally less detailed than the mouth output.
The final stage of the signal chain is a 6db high pass filter, which can indicate size and proximity of the source.
fauna is freeware and can be downloaded here.
2008-02-06: Fauna is now released as donationware. A demo version is available for download (demo has 16 patches, paid version has 107).
More info here.