Skip to main content

How DeepFake actually works

 DeepFake is the process where a program swaps faces with anyone in any video. This A.I. Technology is extremely powerful, it goes through millions of frames to swap faces. Every frame is accurate. The outcome video is very realistic and anyone can be easily fooled by it. These programs use not just one, but two different A.I.s working together at the same time. The first A.I. will go through and scan many images of the subject who is to be faked and then create new faked images. The second A.I. will then examine those faked images and compare them to the real images, and if the second A.I. spots a difference between them, the second A.I. will mark the image as an obvious fake and will pass this information to the first A.I. So the first A.I. takes this information and continuously adjusts the fake images based on the second A.I.’s feedback. Then the new faked images are again passed to the second A.I. for comparison, then again, if the second A.I. spots errors it will share that information with the first A.I. and then again the cycle continues till the point that the second A.I. cannot spot the difference. In simple words; the first A.I. creates fake images, then it shares those faked images to the second A.I. And the second A.I. gives feedback to the first A.I. telling it to redo this image as it is an obvious fake, or telling it to move on to the next image as the second A.I. can't spot the difference.

This system is called Generative Adversarial Network and although the idea behind it is fairly simple, with the ongoing growth of hardware and processing power, it has become a very powerful tool for producing convincing looking fakes. DeepFake also features lots of cool applications. An A.I. network like this can analyse lots of people with a similar look to create a fake model for advertising purposes. It’s way cheaper than to hire an actual person as a model and paying them for a photo shoot. DeepFake could also help with age progression photos to help find missing people, or even to upscale old school pixelated video games so they fit in with the modern displays. But this was simply the tip of the ice-berg. Majority of DeepFake videos right now are pornographic in nature, with the idea of being to swap in a celebrity's face without their consent. Resulting in several actresses falling witness to this activity. Something which also raises a bright red flag is DeepFake’s terrifying ability to match even accurately match up lip movements which raises a concern that it might be used to spread fake videos of politicians saying… well… anything that might get them to the unpleasant side of the public opinion. And when it comes to politics, people often tend to blindly believe almost anything they see.


Comments

  1. Never knew something like this would be terrifying...

    ReplyDelete
  2. "Resulting in several actresses falling witness to this activity." thats criminal

    ReplyDelete
  3. Really glad that you're shedding light on such an important topic which affects us all

    ReplyDelete

Post a Comment

Popular posts from this blog

Times when DeepFake fooled the Internet.

There was a new TikTok channel, obviously we don't know about it because it's banned in India. The TikTok account goes by the name "DeepTomCruise". It got viral. Everyone was shocked, some found it creepy, some  find it cool. There was another viral video made by CorridorCrew, a famous YouTube channel based on VFX. They made a scripted video where they are in a store, and they find Keeanu Reeves in there the title of the video was named "Keanu Reeves stops Robbery." https://www.youtube.com/watch?v=3dBiNGufIJw this is the link to the video, I would recommend watching it. This was a regular humourous video with a ittle bit of VFX. Every one thought it was a regular, everyone was a totally normal video. Few days later CorridorCrew made another video titled "How We Faked Keanu Reeves Stopping a Robbery" https://www.youtube.com/watch?v=IzEFnbZ0Zd4 this video revealed that the video they made few days ago was a not entirely true. Everything was real exce

The Fabricated reality by DeepFake

Throughout the years the art of deception has been commonly used. Fake e-mails, fake text messages, fake images where people's faces are replaced… But as time passed people started seeing through these deceptions. The email address is a clear giveaway, the phone number is a give away. Images are pretty tough to see through as tools like Photoshop would allow one to drastically change the reality of the image to anything they want, but to counter that apps are created which use Deep Learning A.I. to see if the image is Photoshoped or not, and they seem to be doing pretty well. Even without these apps people don't directly believe any image they see on the internet as everyone knows by now how easy it is to fake an image. I mean, it's just an image, Photoshop would easily make something like these. But, it's a totally different story when it's a video… I don't think anyone would doubt that the video is fake. It's easy to do it on Photoshop, you just need the p