This is pretty freakin cool.
A company called VocaliD has developed a system to create more natural sounding synthetic voices for people suffering from severe vocal impairments so that they’ll no longer have to sound like a very limited selection of computers. Using the voices of people who can speak normally, a recording studio and what I’ll bet is some pretty complicated software, they’re able to blend the sounds the recipient of the voice is able to make with elements of the donated speech.
The recording process takes between two and three hours and donors read or repeat short sentences that together cover all the combinations of sounds that occur in the English language. VocaliD is still in its early stages but hopes use a web program or phone app so that people can record their voice remotely in a quiet place in the future.
Using the recording, experts then use software that blends the surrogate’s voice with the sounds of the ‘target talker’ or patient, which creates speech ‘units’ like vowel sounds and consonants.
Once all these units have been made, the synthetic voice is created and allows patients to say any sentence – even those not pre-recorded by the surrogate.
There is a chance that by being a donor, someone could have exactly the same voice as you, but the organisation said it is ‘unlikely’ as the new voice is a blend so will probably only have elements of the donor’s voice.
That’s the short version of how this works. You can find a longer one here.