A legendary US film critic who underwent life-saving cancer surgery has been given his voice back by a team of scientists from Edinburgh.
Roger Ebert lost the ability to speak following a procedure almost four years ago. Since then, the 67-year-old has relied on hand written notes, a basic voice synthesiser and sign language to communicate.
However, now a team of researchers in the capital has used new computer technology to allow him to communicate in his own voice again and Mr Ebert showcased his new communications system in an interview with Oprah Winfrey.
He said: "It still needs improvement, but at least it still sounds like me. In first grade, they said I talked too much. And now I still can."
The firm behind the development, CereProc, painstakingly pieced together snippets of the commentator's recorded work to recreate the exact sound of his natural voice. The team then created a database of words, before creating a prototype which can "speak" any phrase which Mr Ebert types. The system was delivered two days ago.
Chief technical officer Matthew Aylett: "When Roger first heard it, he was very excited because he never thought he'd be able to communicate in his own voice again.
"We're very happy to have been able to use our technology in a way which is helping someone express themselves."
CereProc works with "text-to-speech" technology and specialises in producing voices which have character and emotion. The research team became involved with Mr Ebert after he found out about their work online and contacted them last summer.
Fortunately for the veteran movie critic, his years of TV appearances and numerous DVD commentaries meant the research team had a wealth of audio material to draw upon. The system they have created for him also allows him to tweak the output to correct intonation and emphasis.
Dr Aylett said: "One of the things that we specialise in is trying to produce voices which have got a bit of character and don't sound neutral or boring.
"Conventional speech synthesis normally involves recording the data very carefully in a studio, where we can control what people say and the environment it's recorded in.
"In this case we're using audio that has been recorded for commentaries on Casablanca or Citizen Kane, for example. We have to take this audio and try and produce something which sounds smooth and natural.
"This synthesis sounds very much like a natural voice. It's very important for us that the voices we're trying to reconstruct sound as close as possible to the original speaker."
However, while the team has created natural voices before, this is thought to be the first time they have used specific archive audio for the purpose.
Dr Aylett said: "This is the first time we've actually built a specific voice for someone who needs it.
"Roger Ebert is hoping to use it to produce his blog as an audio piece and he's even thinking he could maybe use it on radio and television."
Dr Aylett said the technology has moved on "dramatically" from the type of speech system developed in the 1980s and famously used by Professor Stephen Hawking.
He said: "Stephen Hawking is now identified by that voice, so he doesn't want to change it. The difference here is the voice we're giving Roger Ebert is actually the same voice he had before he had surgery.
"When he uses it, people who listened to his commentaries in the past as a broadcaster will recognise his voice.
"Both are examples of the way the voice is so central to who we are as people and so much of our character and personality is expressed by the use of our voice."
