Zoomers on TikTok pretend to be the creation of the Veo 3 neural network and collect millions of views

Since the moment when Google introduced its Veo 3 model that can generate realistic videos based on text queries, distinguish between reality from AI creations has become much more complicated — and bloggers from TikTok are taking full advantage of this, pretending to be video versions generated by the technology and collecting millions of views.

Journalist ArsTechnica Kylie Orland investigated the change in the social network’s algorithms after the launch of Veo and found that it is now simply drowning in video fakes. Interestingly, between fake news and surreal short films created by the neural network, TikTok is actively promoting videos with real people pretending to be avatars generated by Veo.

Interestingly, not only teenagers, but also real brands suffer from this kind of «. For example, the indie music band Kongos released a video that they wanted to pass off as the work of AI and even added a promo in the caption: «A band of brothers with beards playing rock music in 6/8 with an accordion». A quick check by Orland revealed that the musicians in the video were a real band called Kongos, and the song (which was quite good, by the way) they played was written 9 years ago.

@kongosmusicI can’t believe how good AI is getting. Pay close attention to the guitar solo at the end and see if you notice it.♬ When You’re Here – In-Studio Live – Kongos

Given how fickle TikTok’s algorithms are, bloggers are resorting to more and more inventive methods of promotion and Google’s Veo has obviously become one of those tools. However, the latter trend looks somewhat disturbing if we look at the popularity of videos where people (allegedly generated by AI) say with sadness, fear, or even paranoia that they are «just tips».

@kiidchroni AI is getting out of hand!! #ai #veo #veo3 #realorfake #prompt #google #gemini ♬ original sound – kiidchroni

So how do you distinguish between reality and fiction?

It’s hard, but it’s possible. There are certain signs that the video is a Veo creation: first and foremost, Google’s model can’t yet create videos longer than 8 seconds, so be aware of the length; in addition, neural network products have a subtle but distinctive style that can distinguish them from regular videos shot on a smartphone — the lighting is too bright, the camera movements are smooth, and the edges of objects are too polished. Alternatively, just look at the channel’s previous videos, and if the same person appeared in «regular» videos two weeks ago, it’s unlikely that they would suddenly appear in Veo’s creations.

All in all, it all looks like a harmless joke for now, but Veo has great prospects of becoming the main tool for dipshits, given that even the simplest tools are repeatedly mislead people on the eve of the US elections or in in the context of the war in Ukraine.