One of the most famous TV movie critic in the country, Roger Ebert lost his lower jaw and ability to speak in 2006. He went under the surgeon’s knife for jaw cancer and has not been able to speak since then due to a tracheostomy – a procedure that includes making an incision in the windpipe. For those of you that need a refresher, Ebert is half of the famous duo, “Siskel & Ebert” that trademarked the thumbs-up or thumbs-down for each movie they critiqued. After Gene Siskels’ death in 1999, Sun-Times reporter Richard Roeper filled the spot.
Since the surgery, Ebert has developed his own methods of communicating, using a small notebook to pen messages and a unique form of sign language where he traces the letters on the palm of his hand with a finger. In a recent interview with Esquire he demonstrated a new piece of technology that gives him a speaking voice.
The generic American voice called Alex, is computer generated and replaces an earlier once called Lawrence. Ebert liked Lawrence because it had an English accent and reminded him of his early summer days in London with his wife, Chaz. The voice became troublesome to understand when it oddly repeated names of people and places.
Edinburgh-based text-to-speech company CereProc is working on a system that will give Ebert his own voice back. The company creates custom voices for people that have lost their speaking ability so they can have a human quality and character to the sounds that are produced from their computers when they type the words. Until now this type of technology created a robotic sound that is evident in the speaking voice of the famous scientist and author, Stephen Hawking.
CereProc currently provides standard voices with regional Southern British, Scottish and American accents, but they can engineer software for anyone that has the fortune of previous lengthy recordings of their voice. The software is compatible with Microsoft Windows PCs and Apple Mac OSX. Voices can be developed in any language. The fact that Ebert was a TV commentator for many years equates to many hours of recordings available for reproducing his voice with authenticity. He is recorded in immaculate digital audio as well on a few (4-5) DVD commentaries.
Apparently English-speaking people use in the range of two thousand words in normal daily activity. CereProc can use Ebert’s recordings to replicate the sound of his voice with a fairly high level of accuracy. They can extract specific syllables and piece them together like a puzzle for words they cannot mine from the footage. By the time the project is complete, Roger Ebert may not sound exactly like he did before the cancer, but it will be miraculously close.