MIT's video morphing technology

At MIT, they can put words in our mouths
The Boston Globe reports that MIT has developed a technology for morphing video images to make subjects appear to say anything. This means a person could be recorded saying one thing, then MIT's new software can create a video of the person saying something else. (Read the original publication in pdf format: Trainable Videorealistic Speech Animation)

As noted in the Globe article, the technology has both positive and negative uses. For foreign language video, it could allow for very convincing over-dubbing. Dead or aging actors could make cameo appearances from their heyday to deliver new lines. On the other hand, it could be used as to create convincing propaganda or to discredit trusted figures.

As artificial video effects become increasingly realistic, authenticity becomes more important. Digital watermarking, a technology currently used to protect copyrighted digital audio files, could be used to verify the authenticity of videos. A watermark, based on a public key encryption system, could act as a digital signature that would be destroyed if the video were modified. Unfortunately, it would be prohibitively difficult for every agency that releases video footage and every television and computer that receives videos to have an encryption key stored on a trusted computer. But it's feasible. I doubt that people care enough or distrust the powers that be so much as to demand such a verification system.

Of course, forged documents are hardly a novelty. Videorealistic speech animation is not some radical new way to mind-fuck the masses. It's an inevitable software application as computers better understand images and sounds. And if it's crucial to verify and maintain the authenticity of a video, cryptographic methods exist to do that.

More security news: Using gelatin to fool fingerprint readers

← Previously: About a Boy | All posts | Next: Free is good →