Even at first glance, there is something about the body on the road. The white sheet under which it is under it is a bit clean, and the movement of the officers is completely empty. “We need to clean the street,” one of them says with a hand gesture, though his lips do not move. That’s fine, okay. But here’s the cooker: My gesture did not include any dialogue.
Google’s new AI video generation model, VEO 3, added the line itself. Over the past 24 hours, I developed a dozen clips, depicting news, disasters and gifly cartoon cats with convincing audio – some of which were invented by the model itself. It’s a little strange and more sophisticated as I thought. And when I don’t think it will now move us forward on the day of the misinformation was lost, but V3 has attacked me as an absolute AI slope machine.
Google introduced VEO 3 this week in I/O, which highlighted its most important new potential: to create a sound to go with your AI video. “We’re entering a new era of creation,” Google’s Gemini VP, Josh Wodord, described it in the keynote, calling it “incredibly realistic”. I was not fully sold, but then, a few days later, I had prepared a video of a news anchor on View 3 that announced a fire in a space injection. All of this was an expensive subscription of the basic text prompt, a few minutes, and Google’s AI Ultra Plan. And you know what? Woodward was not exaggerated. It is like a realistic hell.
I tried a news anchor prompt after watching the clinical instructor Alejandra Karbilo at the Harvard Law School’s cyberla clinic. One of his clips presents a news anchor that announces the death of the US Secretary of Defense Pat Hegsit. She is not dead, but the clip is incredibly convinced. A post that includes a wire of videos that protested with AI infiltrated characters, which protested the indicators used to make them, which have 50,000 opots on Reddate. Scenes in the hospital bed using a breathing tube, a woman, and a character are at risk at gunpoint – all of these spoken dialogue and realistic background sounds. Real lightweight goods!
Maybe i’m bid, but after playing with VEO 3 I am not Enough As I was before. For those who start, clear guards are in their place. You can’t indicate making a video of the Biden tripping and falling. You cannot announce the murder of a news anchor president, or even produces a video of the CEO of the CEO of the T -shirt and a tech company wearing a chain when it receives dollar bills around it. This is a start.
He said, you can create some disturbing mess. Without any smart tasks, I indicated VEO 3 to make a space injection video. Starting with his photo of Mount Renner, I created a video of it, spreading with smoke and lava. In conjunction with a clip of a news anchor, I can see how you can easily seize some adultery with this device.
Here is a good news: It doesn’t look like a ready -made deep fick machine. I gave him a couple of my photos and asked him to make a video with a specific dialogue and will not be complied with. I also said that they should bring a couple in a picture in a picture and drive them out of the scene. It arranged a boot stumping across the sidewalk with some hilarious cranching noise in the background.
I had an easy time to produce videos when my indicators were less specific, which is why I confirmed what I Andrew Marino identified by my colleague Andrew Marino: VEO 3 is the best in creating the lowest general-dinominator YouTube content.
If you have never been hit by an endless garbage pit on YouTube children, let me illuminate you. Imagine that the worst 3D rendering of a monster truck is running under a ramp, descending into a watt of color paint. In front of it, another monster truck takes another ramp into another watt of paint – this time, a different color. Now see it again. And then on YouTube, there are several hours of the luggage, which are small little little little little little little little little little little littleen little little little little little little little little little littleer lords. These videos are generally harmless, designed to prepare only empty calorie views that look like cocomalin Urban cans. In about 10 10 minutes with VEO 3, I collected a clip after the same basic formula – completed with Jonny’s background music. But this clip that is even more disturbing to me is two cartoon cats on a pier.
I thought it would be funny to complain to cats to each other that the fish was not cutting. In just a couple of minutes, I had two cats and a clip with some AI infield dialogue that I never wrote. If it is easy to make a clip of 10 seconds, it would be normal to extend it to a seven -minute YouTube video. In its current form, when you try to expand them in long scenes, clips return to VEO 2, which removes audio. But the way Google is pushing these tools a lot, I can’t imagine that before you can edit a full -length video with VEO 3.
In fact, I am wondering if such a use is a feature for AI-Infiltration Video and not a bug. Google showed us some fancy e -infield video of the real film banners, including Eliza Macent, who is working with Darren Arunofsky in a new film with some AI influx elements. And of course, AI video can prove to be an interesting tool in the right hands. But I think that what we are seeing is the spread of such anxiety that is so good in producing AI – this time, in stereo.
