What I mean is, if the video is a pastiche, then the talking motions ought to be off from the voice, shouldn't they? Maybe the head and voice are him and the body someone else.
Yeah, they take a sample of all his speeches, load it into a database, and the algorithm will analyze it, and interpolate it to new text, and you can then basically make it say anything you want. Any ‘artifacts’ can be manually fixed.
They basically have an actor read the speech, mimicking trumps voice abd movements, and then the deep fake AI just fills it in the details to make it hyper realistic
Abd Audio takes 1/10,000 the computing power of video
What about the sound?
That can also be easily done with machine learning. Deepfakery is just the trendy root word for machine learning.
What I mean is, if the video is a pastiche, then the talking motions ought to be off from the voice, shouldn't they? Maybe the head and voice are him and the body someone else.
Yeah, they take a sample of all his speeches, load it into a database, and the algorithm will analyze it, and interpolate it to new text, and you can then basically make it say anything you want. Any ‘artifacts’ can be manually fixed.
They basically have an actor read the speech, mimicking trumps voice abd movements, and then the deep fake AI just fills it in the details to make it hyper realistic
Abd Audio takes 1/10,000 the computing power of video
If you have enough voice samples from someone, you can fake anything you might want to have them say. Visit https://15.ai/ and try it out.
that's pretty cool
https://youtu.be/oCh27k5nCyI