This article was taken from the June 2014 issue of Wired magazine. Be the first to read Wired's articles in print before they're posted online, and get your hands on loads of additional content by <span class="s1">subscribing online.
Rupal Patel makes voices for people. VocaliD, the Boston speech scientist's project with Tim Bunnell from the University of Delaware, creates personalised vocalisation for people with speech disorders. "I realised that patients using computerised voice devices all sounded the same," says Patel, a director at Northeastern University. "I wanted to do something about it."
The human voice is a combination of vibrations of the voicebox and the vocal tract, which includes the throat and the mouth. In cases of speech disability, patients have difficulty with the fine motor actions responsible for the articulation of speech, such as moving the tongue and lips. But their voicebox remains relatively intact, according to Patel. "Using short samples of speech from a patient, we can evaluate parameters like pitch and voice quality," she says. A "voice donor" then records a three-hour reading of various sentences to extract a bank of speech sounds. "The two are then combined into a single personalised voice," says Patel. "We've only had a few surrogate voices so far, so we can only do a centralised American accent. But we have about 12,000 people who want to donate their voices, many of whom are bilingual. We'll be able to have real variety in the near future."
Patel's team has made about a dozen voices; three are being used by patients. "It took us years to understand what features of the voices we should use when blending them together," says Patel. "We now have a first iteration of the technology and we're ready to scale." Recordings are currently done in a studio, but VocaliD is in the process of developing software to collect surrogate voices remotely. "People usually say that they hate the way they sound.
But what if you lost your voice and had to sound like someone else?
I was talking to girl we made a voice for. She told me that people are finally seeing her for who she really is."
This article was originally published by WIRED UK
