Even at first glance, there’s something off about the body on the street. The white sheet it’s under is a little too clean, and the officers’ movements are totally devoid of purpose. “We need to clear the street,” one of them says with a firm hand gesture, though her lips don’t move. It’s AI, alright. But here’s the kicker: my prompt didn’t include any dialogue.

Veo 3, Google’s new AI video generation model, added that line all on its own. Over the past 24 hours I’ve created a dozen clips depicting news reports, disasters, and goofy cartoon cats with convincing audio – some of which the model invented all on its own. It’s more than a little creepy and way more sophisticated than I had imagined. And while I don’t think it’s going to propel us to a misinformation doomsday just yet, Veo 3 strikes me as an absolute AI slop machine.

Google introduced Veo 3 at I/O this week, highlighting its most important new capability: generating sound to go with your AI video. “We’re entering a new era of creation,” Google’s VP of Gemini, Josh Woodward, explained in the keynote, calling it “incredibly realistic.” I wasn’t completely sold, but then, a few days lat …

Read the full story at The Verge.