ABSTRACT
On February 16, 1931, the New York Times ran a story on a curious development that had just taken place in England: "Synthetic Speech Demonstrated in London: Engineer Creates Voice which Never Existed" read the headline. l The day before, so the article began, "a robot voice spoke for the first time in a darkened room in London ... uttering words which had never passed human lips:' According to the accounts of this event in numerous European papers, a young British physicist named E.A. Humphries was working as a sound engineer for the British International Film Co. when the studio ran into a serious problem. A synchronized sound film (then still quite a novelty) starring Constance Bennett had just been completed in which the name of a rather unsavory criminal character happened to be the same as that of a certain aristocratic British family. This noble clan was either unable or unwilling to countenance the irreducible-even if seemingly paradoxical-polysemy of the proper name (so powerful, perhaps, was the new experience of hearing it actually uttered in the cinema) and threatened a libel suit if "their" name was not excised. As the film had already been shot, however, eliminating it would have involved huge reshooting costs and equally expensive production delays. Consequently, the producers supposedly decided to explore an innovative alternative: unable to get their star back into the studio to simply rerecord and postsynchronize an alternative moniker-the journalistic accounts are uniformly vague as to why-a print of the film was given instead to Humphries, who used his extensive experience as an
This curious artisanal intervention had become possible because the first widely adopted synchronized sound-on-film system-developed and marketed by the Tri-Ergon and the TobisKlangfilm concerns-was an optical recording process. Unlike the earlier Vitaphone system that employed a separate, synchronized soundtrack on phonograph discs, the new optical recording technology translated sound waves via the microphone and a photosensitive selenium cell into patterns oflight that were captured photochemically as tiny graphic traces on a small strip that ran parallel to the celluloid film images.2 "In order to create a synthetic voice;' so Humphries explains, "I had to analyze the sounds I was required to reproduce one by one from the sound tracks of real voices"; having established which wave patterns belonged to which sounds-that is, the graphic sound signatures of all the required phonetic components-Humphries proceeded to combine them into the desired new sequence and then, using a magnifying glass, painstakingly draw them onto a long cardboard strip. After one hundred hours of work this sequence of graphic sound curves was photographed such that it could function as part of the optical film soundtrack and indeed, when played back on a "talkie" projector, according to the journalist who witnessed the demonstration, "slowly and distinctly, with an impeccable English accent, it spoke: 'All-of-a-tremble; it said. That
was an:' But these words-wonderful in their overdetermined thematization of the shiver that their status as unheimlich synthetic speech would provoke-were in a sense more than enough: the idea of a synthetic sound, of a sonic event whose origin was no longer a sounding instrument or human voice, but a graphic trace, had been conclusively transformed from an elusive theoretical fantasy dating back at least as far as Wolfgang von Kempelen's Sprachmaschine of 179 U into what was now a technical reality.
