• Few-Shot Adversarial Learning of Realistic Neural Talking Head Models
    10 replies, posted
https://www.youtube.com/watch?feature=youtu.be&v=p1b5aiTrGzY
This is pretty amazing, those talking heads of famous portraits and paintings were so surreal and creepy to see
Dang, the development of machine learning is so damn quick, its astonishing how far it has come in a decade.
imagine walking through a gallery of famous portraits that just randomly start talking
Every day, we drift further from God's light and closer to real waifus.
When stuff like this becomes more advanced, I wonder how we're going to be able to differentiate from real footage and stuff that was made by machine learning.
Ultimately we won't probably.
So this is like the most asinine thing to get stuck on, but the single question burning brighter in my mind than all others is: Where the heck is it getting the teeth from for single-shot images? Like the Mona Lisa one for example. As she opens her mouth in the synthesized video, she clearly has teeth. Is it taking that from the driver video, and just creatively inserting them into the synthesized video?
they somehow made mona lisa resemble putin even more
We'll make a machine learning algorithm to detect machine learning generated images!
The eyes give it away but it's still really impressive
Sorry, you need to Log In to post a reply to this thread.