Skip to main content

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 perfect angle of the person in the photo and of the person you have to replace the face of. You just crop it out, smoothen the edges, align the face accurately, adjust the size and use Hue / Saturation to change the skin color of the replacement face if necessary. It's as easy… for an image ofcourse. It's nearly impossible to do this for a video. You have to go through hundreds of thousands of frames just to get the face angle right, hundred thousands more to find the perfect ones. Not to mention the time it will take to crop those millions of frames. Even if one were to do all these work it would take them over 5 hours just to face swap a 5 minutes clip. Not to mention the FPS it's recorded in, if recorded in 60FPS they'll have to do all the work twice the ammont. And there is no guarantee that it would be perfect. Doing something like this is sure to leave some sloppyness behind. This is the reason why people neve doubt a video. This has stayed this way untill now… As technology gets better for the good, it gets better for the bad aswell. Last couple of years there has been significant growth in technology. People spending hours to edit one image on Photoshop can now edit them in just a click thanks to the Deep Learning A.I. which has gotten horrifyingly good, everything has become way more efficient then they used to. They have been developing over the years to show flawless results in no time. The day won't be far where A.I. like Bixby, Google Assistant or Siri would be capable of editing photos just the way you want just by telling them to… Literally just by telling them. But these stuff can get ridiculously dangerous. Incomes DeepFake, not as quick as Luminar (One click photo editing app) but it gets the job done as accurately. 

 

Comments

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

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 sha