|
Briefing
|
- Lip Synced Obama Video – University of Washington researchers developed new algorithms that made realistic lip-synced video of former president Barack Obama from audio clips of his past speeches
- How AI Works – Neural network watched millions of frames of Obama videos, learning to form mouth shapes for different sounds, and grafting them on existing video clip
- Potential Applications – Include improving video conferencing technologies, recreating historical figures in virtual reality, and potentially lead to development of algorithms that detect fake videos
- Previous Techniques – Require multiple people saying same sentences over again to get mouth shapes, an expensive and time-consuming process
- Next Step – Includes having algorithms recognize person’s voice and speech patterns from one hour of video instead of 14
|
|
Accelerator
|
|
|
Sector
|
Information Technology
|
|
Organization
|
University of Washington
|
|
Source
|
|
|
Original Publication Date
|
July 11, 2017
|
