Google's VEO 3 AI Videogenerator is the dream of a slop monitor

Even at first glance, the body has something on the street. The white sheet under which it is located is a little too clean, and the movements of the officers are completely countless. “We have to clear the street,” says one of them with a firm hand gesture, although their lips don't move. It's AI, okay. But here is the kicker: my prompt did not contain a dialogue.

VEO 3, Google's new AI video -video model, added this line alone. In the last 24 hours I have created a dozen clips in which news, catastrophes and doofe cartoon cats are depicted with convincing audio – of which some invented the model in themselves. It is more than a little creepy and much more sophisticated than I had imagined. And although I don't believe that it is still driving for a misinformation day -day day, Veo 3 meets me as an absolute AI -Slop machine.

Google introduced the I/A VEO 3 this week and highlighted the most important new function: generate sound for your AI video. “We enter a new era of creation,” said Google's VP of Gemini, Josh Woodward, “incredibly realistic” in the keynote. I was not completely sold, but then, a few days later, I had Veo 3 a video of a news anchor in which a fire was announced at the space needle. All it needed was a basic text request, a few minutes and an expensive subscription to Google's AI Ultra Plan. And do you know what? Woodward did not transfer. It is hellish realistic.

I tried out the prompt prompt after I saw what Alejandra Caraballo, a clinical trainer at the Cyberlaw Clinic of the Harvard Law School, could produce. One of her clips shows a news anchor that announces the death of the US defense minister Pete Hegseth. It is not dead, but the clip is incredibly convincing. A contribution with a number of videos with AI-generated characters that protest against the input requests that are used to create them has 50,000 bullets on Reddit. The scenes include disasters, a woman in a hospital bed with a breath hose and a figure that is threatened with a gun held – everything with spoken dialogue and realistic background noise. Real carefree stuff!

Maybe I'm naive, but after playing around with Veo 3 quite So worried I was at first. The obvious guardrails are available for the beginning. You cannot cause a video of biden trip and fall. You cannot announce a news spokesman for the president's murder or even generate a video of a T-shirt-and-chain-tech company, who laughs while dollar bills rain around him. This is a beginning.

That means you can create a worrying shit. Without clever problems, I prompted VEO 3 to create a video of the Space needle in flames. Starting with my own photo of Mount Rainier, I created a video of it with smoke and lava. In connection with a clip of a news anchor in which the catastrophe is announced, I can see how you can easily sow this tool with this tool.

Here is the better news: there seems to be no prefabricated deep pake computer. I gave him a few photos of myself and asked him to generate a video with a specific dialogue, and it would not keep up. I also asked to bring a few huge boots to life in a photo and leave them out of the scene. It made a boot that stamped over the sidewalk with some strange crunching noises in the background.

I was easier to generate videos when my input requests were less specific. In this way, I confirmed that my colleague Andrew Marino pointed out: VEO 3 is excellent in creating the type of YouTube content with the lowest Den-Den-Den-Den-the content for children.

If you have never been exposed to the endless garbage pit at YouTube children, let me clarify them. Imagine the worst 3D rendering of a monster truck that drives a ramp down and ends up in a tub of colored color. In addition, another monster truck drives another ramp to another color – this time a different color. Take a look at this again. And again. And again. There are hours of this witness on YouTube, which are designed for fascinating toddlers. These videos are usually harmless, only empty calories that are designed in such a way that cocomelone looks like this Citizen Kane. In about 10 minutes with VEO 3 I threw a clip together after the same basic formula – completely with a captivated background music. But the clip, which is even more worrying for me, are the two cartoon cats on a pier.

I thought it would be fun if the cats complain to each other that the fish would not bit. In just a few minutes I had a clip with two cats and a dialogue with AI generated, which I never wrote. If it is so easy to create a 10-second clip, it would be trivial to extend it to a seven-minute YouTube video. In its current form, clips return to VEO 2 when they try to extend them into longer scenes, which removes the audio. But I cannot imagine the way Google is relentlessly promoted these tools that it will take a long time before you can edit a video with VEO 3.

To be honest, I wonder if this type of use for ai-generated videos is a function and not a mistake. Google showed us some unusual video-video video video by real filmmakers, including Eliza McNitt, who works with Darren Aronofsky on a new film with some AI-generated elements. And sure, AI video could be an interesting tool in the right hands. But I think what we see most likely is a spread of the kind of boring pictures that AI can generate so well – this time in stereo.