general posted by

Putting words in their mouths

An absolutely terrifying report from the Boston Globe newspaper which details an MIT project where manipulated video appears to show people saying things that they have not said.

The trick is done by reprocessing the footage to generate new lip movements for an audio soundtrack. It is currently limited to short duration shots where the person is facing the camera without moving, which makes it slightly less scary. The computer samples the face of someone talking for a few minutes and can thereafter match a short audio speech track accurately with the fake video footage.

Of course the potential for abuse in the future is truly terrible, once they have perfected the technology to cope with moving heads, and one assumes, eventually people in motion. The researchers of course claim that it will be a boon for dubbed films and video games. And to cap it all, researcher Tony Ezzat eventually hopes to create a more complex model that will simulate basic emotions as well, presumably to create a more foolproof result. Yikes!

See if you can tell the difference between doctored and un-doctored clips on the site. I couldn’t. [Via Metafilter]

Comments are closed.

comments powered by Disqus

Side Advert

Write For Us

Personnel

Managing Editor:
Nigel Powell

Associate Editor:
Caitlyn Muncy
Associate Editor:
Dan Ferris
Ecological Editor:
Debra Atlas
Technology Editor:
Fritz Effenberger
Asian Editor:
Hu Ping
Reviews Editor:
Kevin Evans

FB Like Box