Voice Demo

Voice Demo

Explore the most natural and diverse text to speech voices on the market!

X Close

26/03/14 XPRIZE launches A.I. competition to deliver the first humanless TED talk.

The race to build a robotic rhetorician.

TED talks, and the speakers who give them, are renowned for polished and witty delivery through which even the most challenging of topics are made accessible. Last week at the TED Conference in Vancouver, TED's Chris Anderson and Peter Diamandis of XPRIZE announced a contest to create an A.I. that could step into the shoes of TED's skilled public speakers and deliver a talk so engaging that it would receive a standing ovation from the audience. This is no small ask, seeing as TED's stage has been held by a major set of big names, from Bill Gates and Julian Assange to James Cameron and Bono.

XPRIZE's A.I. competition has come under criticism from those who claim that creators will be able to achieve superficial success without needing to do anything really groundbreaking with A.I. Murray Shanahan, a roboticist at Imperial College London, suggested that a better test would be to have the robot go to a stranger’s home and make tea, which would test motor control, object recognition and problem solving – “the hallmarks of intelligence.” He may well be right. But it is still far from straightforward to make a robot who can deliver a talk which goes beyond “simply” being coherent, or even insightful, and manages to make a connection with an audience - goodness knows there are humans who can't do as much! Through sheer novelty it may be easy to win people over initially, but it doesn't take very long at all for a human audience to lose interest if the details aren't on point. Try getting your computer's in-built text-to-speech voice to read out a BBC news article and see for yourself that even with excellent content, poor quality speech synthesis is enough to turn an audience off.

CereProc is no stranger to TED, having been at the centre of Roger Ebert's 2011 talk: 'Remaking my voice'. What's more, our text to speech voices have brought life and character to a host of A.I.s and virtual assistants including Swedish social robot Furhat and android stand-up comic RoboJase' for The Gadget Show. A recent write-up in the Telegraph goes into more detail. What makes our voices so well suited to this kind of application is our patented emotional synthesis, and our commitment to continued development of a TTS engine that will produce expressive, characterful voices.

According to the first wave of information about the TED-XPRIZE contest, two A.I.s will be allowed to compete each year with no guarantee that either will be selected as the victor – if neither prevails the competition simply rolls over to the next year. In any case, all of us at CereProc will be very interested to see what comes out of this contest. Who knows, it could well be one of our voices delivering the winning address.