[Computer guy hat on …]
In 1974 the college lab I was associated with had a device called a “Votrax”. It was a 100%solid state speech generator. It was also commercial off-the shelf stuff for the time, but pretty bleeding edge.
It had a vocabulary of phonemes and your program controlled it by sending it a collection of phoneme identifiers with pitch and speed parameters. What came out sounded like Hawking on a very bad day. But it did work.
I don’t recall whether it was truly synthesizing the phonemes from raw waveforms and an envelope algorithm or whether it was working from low-fi samples of human speech. The fidelity was so bad it almost had to be the former.
[Computer guy hat off …]
[Pilot guy hat on …]
The “Pull Up” audio warnings mentioned above come from the Ground Proximity Warning System (“GPWS”) . These were introduced in the early 70s. Most of the airliners of that day had zero computers. The GPWS device even had its own speaker in the cockpit; there was zero integration with other systems. Even on the early systems the voice was a recorded human, not a synthesized voice. I have no idea what technology was used, but a simple miniature tape loop would not surprise me.
Nowadays GPWS, its sucessor EGPWS, and all the other talking subsystems (TCAS, EICAS, TWS, WSDAGS, etc) are all sampled or recorded human voices for fidelity. Obviously the storage requirements for even a thousand word vocabulary at hi-fi are negligible by modern standards.
[Pilot guy hat off …]
Duh. Only now I thought to look for speech synthesis in Wikipedia. They had it on mainframes in the '60s. Of course, they did a lot of things on mainframes that were of little practical use until after microcomputers came out, and that was in the '70s.
The question clearly says “recording” which clearly is not “synthesis” but, as usual, people answer the question they wished was asked, not what actually asked.
While I have no cites I would be very surprised if in the 70s things like brief phone company messages or similar things were not stored in purely electronic memory rather than mechanical electromaginetic memory. (The number you have dialed, 123-5554321, is no longer in service.) Minicomputers were already being used for many things and this application lends itself very well to electronic recording because of the low memory required. I would bet that by 1975 there were already quite a few applications using electronic memory to store audio.
Dammit, maybe I’m hopeless with Google, but boing boing (no idea how they want to be linked to) had an article recently where they took apart an 80s-style ‘speaking’ car alarm, and found it to be based around vinyl technology.
I asked, ‘So when was the first ‘electronic’ voice system developed? By ‘electronic’, I mean an electronically-generated voice and not one stored on tape.’ A little vague, maybe. What I was getting at was an electronically-generated voice that was not stored on a tape or record or disc. The voice could be recorded, but not stored on what I later called a ‘movable medium’; or it could be completely synthetic.
Good link. While my jumping-off point was the B-58 Hustler with its tape system, I didn’t mean to imply that it couldn’t be a building-sized system. This looks closest to what I was looking for:
Up until the late 70’s the phone company used a device made by the Audicron Company (known for their time-and-temperature announcing machines) for those messages. The Audichron had a magnetic drum about two feet in diameter that could record several messages, and playback each on demand by moving the playback head to the appropriate track on the drum. I got this info from my brother, who worked at the telephone company CO for 33 years.