AI: Reanimating the words never spoken

Apr 16, 2018

Artificial Intelligence is making waves across the world and changing the way we live, work and play. Phones may have personal assistants and homes boast smart speakers – but it seems these inventions are just the start of something much more spectacular. We're exploring the potential of this new technology by taking a look at The Times' latest AI project, which has given a voice to John F Kennedy 55 years after his death.

It was back in 1963 that America's 35th president, John F Kennedy, was assassinated on his way to deliver a speech in Texas. It was a speech that would aim to mend the political fences between leading Texan Democratic Party members – but when former marine Lee Harvey Oswald shot him down in his presidential motorcade, these words were left unspoken… until now.

Expert engineers from Edinburgh's text-to-voice technology specialists, CereProc, got together over a painstaking two-month period to bring this little-known speech to life for the very first time. It started with 116,777 sound-units from 831 speeches and radio addresses. Each one was analysed for pitch, energy, and to establish every distinct 'phone' – the subtle difference in sound segments, like the way we use 'W' in both 'word' and 'weapon'. After uploading the best quality recordings to a huge database, AI got to work. The new computer system recognized Kennedy's oratorical style, considering every unique 'umm' and 'ahh' in his delivery to create results that sounded almost indistinguishable from his natural speech.

Over half a century later, millions of people now have the chance to hear what Kennedy had to say on that fateful Friday. It's a mind-blowing, spine-tingling experience to listen to a voice brought to life from beyond the grave – but more importantly, it's a great reminder of Kennedy's incredible skills as an orator, and why he made such a lasting impression on the masses. 

quote

Kennedy spoke powerfully and passionately on issues that are still remarkably relevant today, and AI has allowed us to mark this important part of history and make it available for generations to come.

But this is just one way this voice-cloning technology can be utilised. The tools used in the JFK project could also be harnessed for those with debilitative conditions like motor neurone disease. Future sufferers will maintain their independence by recording their speech to recreate it further down the line – preserving a vital aspect of their personality when so many are taken by It was an opportunity offered by CereProc to the late Stephen Hawking, but decades communicating with his famous robotic tone made this such an integral part of his identity he turned the opportunity down.

Whether AI technology used to recreate history that never was, or give people a voice when it's cruelly taken away, it's clear that is carries huge potential for changing the way we learn, speak and interact.