This week, a Brazilian content creator named ederxavier3d published an amazing video on his Instagram account. In it they appeared several of the protagonists of the series ‘Stranger Things’ making just the gestures and expressions that he made in the lower window. He explained that he had achieved it in a simple way and thanks to the new Kling 2.6 and its characteristic Motion Control.
This option allows the movements we make in a video to be transferred to any real or fictitious person, regardless of the style in which they are represented (it can be a comic book character) with an amazing result.
And in the case of the Brazilian creator’s video it can be clearly seen: these videos could perfectly pass for real in almost all cases. It is true that if you look closely you can sense that something is wrong, but only because the video has been shared making it clear that it was created with AI.
This video has once again awakened a debate that has been linked to this type of deepfakes for some time. Justine Moore, partner at investment firm a16z, explained that “we are not prepared for how quickly (video) production flows are going to change with AI. Some of the latest video models have immediate implications for Hollywood. Endless character swaps at negligible costs.”
How the story has changed. In April 2023 we proposed a little game to all of you who read us: Would you be able to distinguish a real image from one generated by AI? At that time, AI was already achieving remarkable results—the image of the Pope with the coat proved it—but the feeling was that we could still tell whether an image had been created with an AI or not. With the video things were even clearer.because at that time AI video generation I was in diapers.
Three years later things are very different, and there are several AI platforms (I see 3, Sora 2Kling, Runway) that generate videos that anyone could easily confuse without problems. Tell Will Smith.
ederxavier3d’s video also demonstrates this, and in fact something unique happened with it: there were rumors that the Stranger Things characters had actually recorded those appearances and the creator had imitated their movements, and then rumors appeared that denied that this was true and that suggested that from the beginning the videos were nothing more than what they seemed: deepfakes created with AI.
A priori everything would suggest that this is the case: the Kling 2.6 feature is not at all new and other platforms allow our movements and gestures – and even our voice – to be transferred to a character generated by AI. The problem is that At this point it is almost impossible to distinguish whether that person who appears on the screen is real or not.
This technology is extraordinarily striking and causes that “wow effect” that AI companies so seek, but despite the creative options it offers, the risks here seem especially notable.
The identity theft It is now easier to achieve than ever, and that will probably mean that we will see many more dangerous cases. It is enough to remind the employee that transferred 25 million dollars believing that the person who told him via videoconference was his real boss.
It is not clear how, for example, Hollywood studios will react to this technology, but for now some are already taking action on the matter to try to protect themselves.
The best example is Matthew McConaughey, who these days has “patented himself” to have one more legal resource (which we are not sure is necessary) to protect yourself against possible videos impersonating you. The implications are enormous, and we are entering an era in which something disturbing is going to happen: We will not be able to trust what we see on a screen.
GIPHY App Key not set. Please check settings