SIGSALY was a highly classified method of coding, one-time pad encrypting, and decoding voice transmissions during World War II. Churchill used it to communicate with the US and Allied commanders during wartime.
I never managed to find any recording online of what it actually sounded like. I'd love to hear it.
"A recording of Churchill talking to Roosevelt is on display in the British Parliament Building. It is very interesting to hear since the timing of the recorded signal occasionally varies from that received. The spoken words were added to a very loud Beethoven recording which can occasionally be heard."
.. the bellows-operated "acoustic-mechanical speech machine" of Wolfgang von Kempelen of Pressburg, Hungary, described in a 1791 paper. This machine added models of the tongue and lips, enabling it to produce consonants as well as vowels. In 1837, Charles Wheatstone produced a "speaking machine" based on von Kempelen's design, and in 1846, Joseph Faber exhibited the "Euphonia". In 1923 Paget resurrected Wheatstone's design.
I've contemplated recreating this machine, in part because I'd like to "play" it. There is plenty of material on the design and workings of the device but there is precious little material on how the operators (referred to as "girls" in the demo) were trained, a process that took months.
If anyone has access to such training material, please contact me: The_Amazing_Dr_T (remove underlines) at gmail dot com
I really enjoyed the video linked in the article. I wasn't sure what to expect, I found the Voder "voice" slightly creepy. Sadly I could not find an emulator or tool online to input text and get a Voder audio output.
> Sadly I could not find an emulator or tool online to input text and get a Voder audio output.
It seems like you'd need samples of the individual sound channels from the machine, and then a program to map the text to timed signals to produce the different sounds, along with pitch shifts, etc. Something similar to MIDI, maybe (vibrato, pitch shift, and the different sounds as different instrument numbers seems like it would work).
Part of the charm of the machine is the control the operator has to give inflections, tone, and timing. I think a text-to-voder emulator would miss a lot of the variation heard in the demo and would instead sound stale and less lifelike.
It's interesting how lifelike it sounded (comparatively speaking) and how much better it sounds than early electronic speech systems (like the Speak&Spell or the C64 Sam https://www.youtube.com/watch?v=Y0uexj4OsA4)
The title comes from a difficulty in synthesizing the phrase "How to Recognize Speech" and having it come out "How to Wreck a Nice Beach". :)