Today you can download software that uses machine learning to perform a convincing digital video face-swap. The video is called a deepfake. It can stitch the eyes of one person onto the face of another person doing lip-sync karaoke. It isn’t perfect yet, but it could fool some people.
Photo fakery is certainly not new. It’s been around for decades. But artificial intelligence completely changes the game. AI makes it possible for anyone with a decent computer and some time to do the same thing. You can use free, open source software called OpenFaceSwap. It’s available from Deepfakes.club.
And this is just the beginning of powerful fakes. As machine-learning advances, the fakes will become much harder to spot. This has potential to create great harm to our society. AI doesn’t just give us a better Photoshop or Premiere, it helps the computer learn what the world looks and sounds like which can then be emulated.
To do the work, you will need an advanced graphics chip in you computer. These chips cost a few thousand dollars. The good news is that you can rent access a virtual machine for pennies a minute using a cloud machine-learning platform like Paperspace.
So when you get this far, the next step is easy. Just feed in two video clips where you want to put one face on the other and go have dinner while the program learns how each face looks and moves so that it can map one onto the other. The two clips you choose must be similar to each other. If they are too different, the results will be a mess.
Recent Comments