This week, researchers from the Samsung AI Center in Moscow released a paper called “Few-Shot Adversarial Learning of Realistic Neural Talking Head Models” outlining how they were able to create moving animations of a person based on just one source image. This enabled them to bring a number of paintings to life in a realistic way, including the Mona Lisa.
These new deepfake technologies can be really creepy, but they also provide new ways of bringing old and historic things to life, much like I’m trying to do with the Historic Voicebot. Maybe in a few years time we’ll all be using deepfake AIs instead of cut out animations, who knows?
Mona Lisa ‘brought to life’ with deepfake AI, BBC Technology, via https://www.bbc.com/news/technology-48395521 Date of reference: 24th of May 2019.
It’s Getting Way Too Easy to Create Fake Videos of People’s Faces, Samantha Cole, Vice, via https://www.vice.com/en_us/article/qv7zkw/create-fake-videos-of-faces-samsung-ai-labs-algorithm Date of reference: 24th of May 2019.