“We can always grab characteristics of their voice and reapply the process as they go along,” he said. “With individuals who have neuro-degenerative diseases such as ALS, we capture speech from them right after they’re diagnosed, while they’re still speaking fluently and create their new voice from that. It captures elements of their voice that default devices don’t typically use.”
VocaliD was developed as an extension of another product called Model Talker, a text-to-speech synthesis system developed by Bunnell. A goal of VocaliD is to personalize the speech so that not only is the voice unique, but all vowels and consonants are clearly understandable.
Right now, although the technology is already developed, it is primarily a windows-based software product. Further development will enable the product to be used on I-devices (I-Phone, I-Pad).
“Making that happen is really more of an engineering kind of feat,” said Patel, who is cross-appointed in speech and language pathology, computer science and engineering. “What we’re doing here is manipulating the voice. There are no other labs doing this kind of work, although there are other efforts in trying to build voices. But they are categorically different because we are using a small amount of speech from our target talker.”
In addition to creating voices for speechless individuals, Bunnell sees other long term applications for VocaliD.
“It has lots of possible applications,” he said. “Those include things like rapidly developing multiple voices for games, or voices for reading books or novels. Right now the quality of the voices is still variable. Sometimes we are able to produce very good sounding voices and sometimes we can’t. We are still trying to work that out. This is an ongoing research project.”
Ultimately, users will likely be able to communicate in their own unique voices, from mobile communication devices as compact as a cell phone.