How Deepfake algorithm works | Ian Sullivan?

Deepfake algorithms make it look like someone is saying words they never actually said, like a movie star talking to you without them even being there.

Imagine you're playing with your favorite toy, and you want your friend's face to appear on the toy. You take a picture of your friend, then use a special tool that knows how their face moves when they talk. It copies those movements and puts them on the toy, making it look like your friend is talking to you from the toy!

That’s what deepfake algorithms do, but with videos instead of toys.

How It Works

  1. The algorithm starts by looking at lots of pictures or videos of a person’s face.
  2. It learns how their face moves when they smile, frown, or talk.
  3. Then, it takes another video, maybe someone else talking, and changes the face in that video to match the person it learned about.
  4. The result is like watching your friend speak on a screen, even though they’re not really there.

It’s like giving your toy a new voice and face, but with real people!

Take the quiz →

Examples

  1. A person's face is replaced with another in a video, like making a politician say something they didn't actually say.
  2. A deepfake uses photos to make someone look like they're saying or doing things they never did.
  3. You can create a fake video by training an AI on pictures of a person's face and then using it to generate new expressions.

Ask a question

See also

Discussion

Recent activity