A Nixon Deepfake, a 'Moon Disaster' Speech and an Information Ecosystem at Risk

Abstraction: MIT-created Nixon deepfake exposes synthetic media misinformation risks

Key points:

• MIT artists Francesca Panetta and Halsey Burgund collaborated with Canny AI and Respeecher to create a posthumous deepfake of Nixon delivering an undelivered "Moon Disaster" speech
• Canny AI used "video dialogue replacement" (not face-swap) to transfer facial movements to a Nixon reconstruction; Respeecher used voice conversion to synthesize Nixon's voice from an actor's reading
• The project took more than half a year to produce; the full process was documented intentionally
• 96% of deepfake videos online in 2019 were non-consensual pornography, per Deeptrace Labs
• Deepfake techniques include face-swap, audio fakes, lip-syncing, and whole-body puppetry — some already used commercially in Hollywood and medical applications
• The MIT team worked with Scientific American to expose the technology's implications for history and the information ecosystem

Connections: Mit · Canny AI · Respeecher · Deepfakes · Synthetic Media · Misinformation

Source: https://www.scientificamerican.com/article/a-nixon-deepfake-a-moon-disaster-speech-and-an-information-ecosystem-at-risk1/