The simulation of voice has a long history, which has crossed very different techniques for simulating our vocal apparatus. From the purely mechanical, such as the Euphonia allegedly created in 1820 by the Austrian inventor Joseph Faber, using a piano, a bellows and a replica of the human throat and vocal organs, to current LLM-based software simulating the reading of a text with a wide range of possible voices and accents, and even modifying a speaking voice with specific accents in real time. Among current experimental practices (the whole Neural #78 “Voices, humans, machines” was dedicated to them), there are some ingenious approaches, sometimes inspired by past attempts, which go beyond this kind of simulation, producing a voice with alternative elements. Paul Magee has assembled an imaginative installation with specific mechanisms. “Signal” (which has been part of the Perma/Soils exhibition at the Piksel 25 festival) is made of a matrix of signal generator modules, which modulate sine waves matching human voice frequencies. There are dials on each box, which slowly rotate back and forth to the frequencies they are emitting, and their combination produces a “voice” that can be heard. Each voice uses a set of four generators, and each of them is based on a different band of frequencies, from the lowest at 127 Hz to the highest at 212 Hz and above. The set hosts sixteen boxes, which constitute four separate voices. The analogue nature of the system does not permit perfectly coordinated emissions of frequencies at the programmed time, so the results are not interpretable. Still, we recognise a mechanical voice, the basic quality of a modulated sound that activates our perception of communication. This clumsiness of the system shows the machine trying to catch up not only with the physics it is based upon, but also with the organic nature of our voice, as a live, ever-changing entity that has a very special encoded quality in itself.

 

Paul Magee – Signal