In this blog post, I will generate some deepfakes. What is a deepfake? According to ChatGPT
A deepfake is a synthetic media creation, usually in the form of a video, audio, or photo, generated by artificial intelligence to convincingly alter content, often by replacing a person's likeness or voice.
How to generate a deepfake? There are many online tools that can do this. There are also some projects on GitHub that can be downloaded, allowing deepfakes to be generated locally on your computer. Usually, online tools are paid, but there is a possibility to generate a demo deepfake for free.
I will be using Vidnoz AI Tools for deepfake generation
particularly Photo Face Swap feature
and Video Face Swap
I found Vidnoz really handy, and it is indeed free. In the free plan, the video/photo generation just takes more time. Photo generation works even without registration. I encourage everyone to make some generation experiments for fun.
The aim of Vidnoz photo/video generation is for
business, social media, entertainment, etc.
so I won’t be generating professional deepfakes. Let’s refer to them as ‘light’ fakes instead.
Let’s see Vidnoz AI in action. I uploaded two images to explore the results of face generation.
The first image is of my colleague Julia from Alten, and the second is a selfie of myself.
Depending on the order of Base and Target images, the results are
Let’s now move to video face swapping. For this demonstration, we will use my face and that of Kamil, my colleague from Alten.
Let’s watch the output video from Vidnoz.
On the left-hand side, you can see the original video recorded by me. On the right-hand side is the video generated by Vidnoz. I’ve joined them together for better visibility. As you can see, when an object like a ball obstructs the face, Vidnoz’s generation struggles, and it becomes easy to spot that the video is generated. I hope this convinces you that I am not presenting professional deepfakes, but rather amateur light fakes.