A program called Lyrebird promises to create a digital voice using your own voice, with just one minute of recording you. The startup company lists speech synthesis for people who have lost their voice as one of the possible applications of the program.
The program, named after the Australian Lyrebird that is known for its ability to mimic any natural or artificial sounds of its surroundings, has some demo records on its website to show potential customers, what Lyrebird is capable of. The demos even include various sentences spoken by Donald Trump and Barack Obama, which have been created from the audio data from the former presidents’ past public speeches.
Even though the demo phrases have actually been spoken by the two presidents in the past, it shows a glimpse into potential future problems of the Lyrebird system. The voice recognition features and digital voice replication is such an exact copy of the original, that anyone listening to it, could think it is the real Donald Trump talking. In theory, the artificial voice could be used by anyone in an attempt to impersonate or bypass voice identification features. The company has already highlighted this problem on their website, noting that the recordings are not necessarily trustworthy. In fact, the Lyrebird program might call into question the role of one’s voice as strong evidence, particularly in court cases or it will even lead to deniability of any voice recording.
The company has already implemented a digital watermark to every voice recorded to stop any forgery. However, critics already point out how a misuse of such technology could lead to horrific political outcomes, if for example, the spoken words are potential threats of war or other powerful statements seemingly spoken by a well-known politician.
Still, the importance and impact of the new voice technology has already been recognized by other companies, such as Google and Baidu.
It remains to be seen what the companies actually do to minimize harm for their clients.