Crossing the Uncanny Valley of Conversational Voice

About 6 months ago, I created a video that won the September Club Documentary Competition (Move 37: Dreaming of Electric Sheep). Although it sounded like me narrating the story, it was actually an 11Labs clone of my voice.

I simply typed the words for my script into a text document, and got 11labs to generate the speech. I didn’t need a microphone or recorder to make the narration. Whilst that was stunningly realistic – and many people present didn’t realise it wasn’t a genuine recording of my voice – it did suffer a little bit in that the tone and inflection was slightly flatter than my real voice.

I recall that one member of the audience at the time proclaimed that Ai would never replace a human actors voice in a film.

It’s perhaps unwise to use the word ‘never’ in relation to Ai, since we are now even closer to the point where an Ai voice might be indistinguishable from that of a human in an ordinary real-time conversation.

This demo conversational voice system from Sesame.com shows how close we are today. This is from their website;

They go on to explain:

While it’s great to read the words, the real test is whether it actually sounds as good as their claim? Try it for yourself!  Click the images or Open the link below:

https://www.sesame.com/research/crossing_the_uncanny_valley_of_voice#demo

There is no need for registration (you can skip if asked), just click ‘Try Our Demo’, and choose either Maya or Miles to start a real time conversation.

It’s not a search engine, but has been trained an a large amount of literature, so chat to it in the same way you might talk to a colleague.

I don’t know how long this will be available for free, so try it soon!

Welcome to the future!