"Before the age of electronics, scientists were hindered (in terms of our technology) in that they had no transistors or integrated circuits to fill their talking boxes. They, instead, had to rely on the sceince of mechanisms that they knew. Much like Charles Babbage's room-sized analytical machine "computer," constructed solely of mechnical parts, the early talking machines were more like musical instruments than the complex systems we might expect. The mechanical technology used at that time should not be taken lightly. Research efforts at constructing mechanical voice systems are still ongoing in parallel development with electronic synthesis methods.
One of the first attempts at mechanical speech generation in recorded history occurred around 1770 at the Imperial Academy of St. Petersburg. In response to the university's challenge to explain the physiological speech differences between five vowels, Kratzenstein won the annual prize for modeling and constructing a series of acoustic resonators patterned after the human vocal tract. The speaking device, crude by today's standards did, however, create the vowel sounds with vibrating reeds activated by air passage over them. By varying the acoustic resonators and effectively selecting the formant frequencies by hand, limited speech was generated by this mechanical device.
Unknown to Kratzenstein at the time, Wolfgang von Kempelen was working in parallel on a more elaborate speaking machine for generating connected speech."
John P. Cater, "Electronically Speaking: Computer Speech Generation", Howard M. Sams & Co., 1983, 72.
Kratztenstein's resonators for synthesis of vowel sounds.
The resonators are actuated by blowing through a free,
vibrating reed into the lower end.