City firm puts the words back into famed film critic's mouth

Published on Thursday 18 February 2010 09:01

ONE of world's most famous film critics has been given his voice back, thanks to a team based in Edinburgh.

Roger Ebert, 67, better known as one half of the renowned US critics Siskel and Ebert, lost his voice to throat cancer four years ago, during operations to save his life.

Since then he has communicated by writing words on notepads or using a basic electronic voice simulator similar to the one made famous by Professor Stephen Hawking.

Now however, he is set to get his own voice back – or at least a simulation of it – thanks to the work of experts in the Capital.

An early version of the voice is set to be heard live by millions of people tomorrow when Ebert is interviewed by Oprah Winfrey on her chat-show.

It will be an emotional moment for Ebert, marking the first time he has 'spoken' in public for years, and it will be an equally momentous occasion for the dozen staff at CereProc who are still working to perfect their version of his voice.

Matthew Aylett, the Chief Technical Officer of CereProc, said: "We are still working on it, but we have given him an early version which he will be using on Oprah. I'm a bit nervous, but I'm sure it will all go fine."

Set up in 2006, the company normally creates entirely synthetic computer voices, with Scottish accents, which can be downloaded by people for their computers.

Their speciality, and the reason they were contacted by Ebert, is putting emotion into their voices, to make them sound more human than ever before.

They have already landed major contracts to produce an in-car voice for Ford – which they were told should be able to tell jokes.

They even created a synthetic version of former US President George W Bush's distinctive voice for a satirical website, which allowed users to literally put words in his mouth.

Thanks to the vast number of sound recordings of Mr Ebert taken from his TV reviews as well as DVD commentaries he carried out before losing his voice, they have been able to painstakingly reconstruct his voice.

HOW IT WORKS

THE technique used by CereProc has seen the firm go through thousands of hours of audio recordings to pull out key words spoken by Roger Ebert.

As well as the database of words, they also use certain sounds or snippets from parts of words to help them construct a library which allows the user to say anything they want.

The audio files can also be manipulated and tweaked by the user themselves, to help express emotions.

It means they are able to make their voice sound angry, happy, or stressed.