On August 28, a video of a young Kylian Mbappé being scolded by a man said to be his father started circulating on social media. Initially published by a satirical account specialising in comedic "deepfakes" – artificially-generated images that replace one face with another – the video was reposted multiple times with some people taking it seriously. The creator of the video told us more.
A "deepfake" video of Kylian Mbappé went viral on various social media platforms. Although it began as satire, the video was shared and altered multiple times, to the point that it lost its original context.
"I'm going to deal with your problem. I'm going to give you a good beating because it seems to me that when someone talks to you, you don't understand, you are struggling," a man says to a boy in what appears to be a school.
The video, published on August 28 by the Twitter and TikTok account French AI Covers (@frenchaicovers) has the caption "FLASH - Exclusive Video: Wilfrid Mbappé reprimands Kylian Mbappé for his wavering feelings regarding his future at PSG". The caption refers to the French footballer's recent hesitation between extending his contract with Paris Saint-Germain and joining Real Madrid. The child seen in the video closely resembles a young Kylian Mbappé.
However, the video is a "deepfake," a video generated using artificial intelligence in which the face of the child actually filmed in this sequence has been replaced by that of Kylian Mbappé.
‘It only took me a few hours to create this video’
The FRANCE 24 Observers team spoke with Younes, a 25-year-old computer developer and the administrator of the @frenchaicovers account:
I've been passionate about computer science since I was very young, and I'm highly interested in 'deepfake' technology. I taught myself by searching for free tutorials on the internet that allowed me to easily create a 'deepfake'. It mainly requires having a good graphics card but it doesn't require any particular expertise, just some ease with open-source software.
It only took me a few hours to create this video of Mbappé from the original footage. I had to make a few editing adjustments to make the face even more realistic, but I'm not a professional editor. I'm quite surprised by the quality of the initial result, which automatically takes into account lighting and facial movements.
What's very surprising is that I trained the algorithm using images of Mbappé as an adult, not as a child. However, the result still resembles a young Mbappé.
Younes shared the production process with our editorial team. After an initial computing phase using software to switch the face of the boy with that of Kylian Mbappé, some modifications were necessary to make the video even more realistic.
‘I've seen it in English, Arabic, and even Spanish’
However, the man behind @frenchaicovers wanted to be very clear about his intentions:
In the context of my Twitter or TikTok accounts, my intention is clearly comedic. It was a joke, and overall, everyone who comments and follows the account, because they know me, understand that it was a parody. I made it as clear as possible by adding a watermark to the video and even by posting a second tweet to clarify that it was a parody.
Younes continues:
The problem is that once it's on social media, I don't necessarily have control over the video anymore. I saw many people grab it, remove the watermark, and post the video with a caption stating that it was young Mbappé. The video has also circulated widely internationally: I've seen it in English, Arabic and even Spanish.
I believe people wanted to create a narrative around this video to make it appear real when, in the beginning, it was just a simple joke. I think it's because people are always searching for new videos from Mbappé's past, videos that haven’t been made public...
This highlights the dangers of ‘deepfakes’: it's a technology that is becoming increasingly accessible, requiring less and less computational power, and is becoming more widespread. I think that today, awareness about ‘deepfakes’ is clearly insufficient.
If you're interested in the subject of "deepfakes" and want to learn how to detect them, you can check out our verification guide article by clicking on the link below.
