How to change somebody’s life by speaking 3,500 sentences

As part of the Fast Company Innovation Festival in New York, I spent this morning learning about the future of speech technology at a session hosted at the offices of Saatchi & Saatchi and featuring that company’s Brad Soulas and the founder of VocaliD, Rupal Patel. But the most remarkable thing they discussed wasn’t some far-off fantasy, but rather what VocaliD is already doing.

The company’s technology allows speech-impaired people to communicate through a synthesized voice tailored to themselves–so that a little girl sounds like a little girl, and a 50-year-old man sounds like a 50-year-old man. It accomplishes this by generating digital voices based on audio recordings by real people, a task that requires individuals to record 3,500 sentences that are then broken apart and reassembled into the sounds needed to form any word.

As part of VocaliD’s Human Voicebank project, over 14,000 people have donated their time to create these recordings. As for Saatchi & Saatchi, it created an animated fairy tale to explain this whole concept, starring a character named Goldivox.