In March 1951 three men huddled around a portable tape recorder. One of the men, Mr. Osmond “Ken” Kendall of the National Film Board, pressed play. From the small speakers came the voice of Brooke Claxton, the minister of national defence. Claxton had recently announced Canada’s rearmament programme in a press conference, and the recording seemed to be a broadcast of the minister’s remarks.
But there was something different in this version of the speech.
The other two listeners, officers of the Department of External Affairs (DEA), heard Claxton speaking pessimistically about Canada’s ability to rearm, and the impossibility of Canada meeting its force increases by voluntary enlistment. It was ostensibly Claxton’s voice, but the Minister’s message was precisely the opposite of what he’d said in his speech and the following press conference. The recording manipulated and distorted the original.
It was an audio deepfake.
Kendall had played his recording for George Glazebrook and Jack McCordick, the men in External Affairs with special responsibility, respectively, for intelligence and psychological warfare. Kendall brought him with the diagrams of the machine and its electrical circuits:
The machine takes a tape, on which is recorded the speech being handled, and plays it in such a way that the words and syllables can all be precisely identified on the tape. At the same time, the machine plays a second clear tape on which the new version is recorded. The new version is worked out before the machine is stared and the original tape is suitable marked. It is then possible, through the system of marking, and through the electronic and mechanical interlocking of the two sides of the machine, plus the instantaneous braking of either side, to rearrange the original speech according to the pattern decided upon.
Anyone who grew up making mixtapes in the ‘80s or ‘90s will understand the basic principles at play. But for the DEA officers in 1951, it was a surprise that one might manipulate a recording so easily – and convincingly. Kendall showed them that it was simple to delete negative words, flipping the meaning of remarks; or indeed to rearrange phrases and words entirely. An existing speech could simply be altered. But there was opportunity for even more modification: A trained speaker — what we might call a voice actor — could record remarks that would then be inserted into an existing recording of a speech to dramatically alter its meaning. The men’s imagination ran to using this method to impersonate dictators past and present. Such a manipulated recording, they thought, might not have fooled German listeners “thoroughly familiar with Hitler’s voice and ranting style.” But the leader of the USSR spoke infrequently and in a monotonous style, few truly recognized his voice. The Canadians thought “it might be quite easy” to deceive a Soviet audience with a distorted speech by Stalin.
Kendall had prepared his example by a “scizzors [sic] and paste” method of manual recording. At the time of the meeting, he had designed (if not yet built) a machine that could perform these tasks quickly and easily. McCormick thought such a device held “great potential importance” and wanted Kendall to build a prototype and begin “a programme of familiarization and experiment.” The correspondence in the DEA files outline a number of possible first steps, such as transferring Kendall from the National Film Board to the National Research Council where he could get the funds to work on this project in secret. Such a step could only follow a security check as to Kendall’s “reliability.” It might then be time to bring officials from the Department of National Defence into the picture. And after the security checks were complete, McCordick planned to give Kendall a set of working papers he’d prepared on the Russian language, materials that might be useful for “considering the distortion of a speech in Russian.”
Glazebrook and McCormick asked Kendall to prepare a further demonstration. They suggested he contact the CBC and obtain a recording of a speech by Secretary of State for External Affairs Lester Pearson, and then prepare three distinct recordings: the correct or original version, a second version only slightly distorted, and a third version with “as little possible resemblance to the original” as Kendall could manage. (Whether or not Kendall proceeded with this request is unclear, but Pearson was apprised of the plan.)
The most intriguing part of the discussion regarding Kendall’s demonstration was the use to which the Canadians thought they might put these audio impersonations and the possibilities glossed over. Although Kendall had distorted Claxton’s speech and was asked to do the same to Pearson, there is no mention of the possibility that these altered speeches could be used to sow discord and confusion at home; no idea this technology could turned against Canada. Quite the opposite: Kendall and McCordick imagined a “mobile unit” of trucks carrying generators, antennae, and crew quarters. The unit could approach an “enemy area” and use its more powerful transmitters to “swamp” the enemy’s own broadcasts, replacing genuine transmissions with an altered speech playing on the same frequencies used by the enemy. The “practical application” of these audio deepfakes would be an offensive weapon of psychological warfare in the Cold War.
In 1955, Maclean’s magazine published an article “Osmond Kendall’s Marvelous Music Machine,” featuring a device Kendall had created that “turns doodles” of a grease pencil “into the sounds of an eighty-piece orchestra.” In true 1950s style, it was named the “composer-tron.” The Maclean’s article went on to note that Kendall had also designed a second version of the composer-tron, built by the Canadian Marconi Company and housed at the National Film Board’s Ottawa studios: a “speech machine.” Did Kendall’s invention end up playing the Cold War role he envisioned, and for which McCordick hoped? More to come…
**
The sources for the above are Memorandum for the Minister from Heeney, March 28, 1951, and Memorandum for the Under-Secretary from MacKay, April 17, 1951. Both can be found in the “Psychological Warfare” files of the Department of External Affairs, Record Group 25-A-3-b, file 50182-40, Library and Archives Canada. The Library and Archives Canada catalogue indicates that LAC holds another file regarding Kendall’s machine. This file is the subject of an ongoing Access to Information request.