What's that, Adobe? A Photoshop for faking voices?

Why Stephen Fry never needs to open his mouth again

His master's voice

Recorded voice evidence will never quite be the same again. It might not even be "evidence".

Last week, Adobe demonstrated a voice manipulation package called VoCo – a research project, and not (yet) a product, it says.

Described as a "Photoshop for voiceovers", VoCo is actually quite brilliant. Given a speaking sample of about 20 minutes, it builds a voice model from the speaker's phonemes. This can then be used to make the "speaker" say anything, using text-to-speech software. Type it in, and you have a ready-made celebrity for use in voiceovers, dedicated greetings and so on.

But there may be problems. The BBC unearthed a media theorist, Dr Eddy Borges Rey of the University of Stirling, who was “horrified by the development".

"This makes it hard for lawyers, journalists, and other professionals who use digital media as evidence," said Dr Borges Rey.

He may have a point. Adobe claims watermarking technology will be used to determine the authenticity of the records offered as evidence, but that raises a huge problem. Will a voice recording offered as proof – say, of Sam Allardyce being indiscreet, to use a recent example – now be discounted because it isn't watermarked? No doubt there are technical solutions to technical problems, but many create even more.

Here is the goofy demo at last week's Adobe MAX 2016:

Youtube Video

We suspect this technology is bad news for soundalike voice artistes – and may prompt a shift for owners of well-known voices such as Stephen Fry (pictured below) to move to an IP-based, licensing model.

But Mr Fry will be able to do a voiceover without leaving his bed – or even opening his mouth. ®

Sponsored: Minds Mastering Machines - Call for papers now open


Biting the hand that feeds IT © 1998–2018