Google's video generation model is almost ready for prime time

Many people’s introduction to generative AI was seeing the uncanny images produced by early DALL-E. Nowadays, companies are focusing on bringing AI-generated videos. Google showcased its video generation platform called Veo at Google I/O earlier this year, and it’s now available as a preview.

Google has been trying to undo the damage done with its Imagen image generation platform. The company stopped generating images of human beings, and it still doesn’t allow this functionality for most people. So, it seems odd that the company is going ahead with a video generation model.

Google released Veo in a private preview

We haven’t seen anything from OpenAI’s Sora since the big leak that happened not too long ago. The clips that we saw from that model were unsettlingly good. We’re not sure when Sora will officially launch, but we’re sure that it will be before Amazon’s Nova Reels model.

Before those models will be free to put filmmakers on the street,  Google released its video generation model. Just know that most people won’t be able to use it. Google has only released a private preview of the model. You can use it if you’re subscribed to the company’s Vertex cloud service. We don’t know how long it’s going to remain in the preview state. Knowing Google, we’ll see some sort of public release down the road. The company hasn’t been shy about giving users free access to its tools.

At the moment, we don’t know how long the video clips can be. However, we know that Veo spits out videos at 1080p resolution.

So, how good are the videos?

Just like other videos we’ve seen generated with AI, the results are extremely realistic. They’re a far cry from those videos of Will Smith eating spaghetti, that’s for sure. They’re to the point where you’d have to heavily scrutinize the videos to notice any tells.

For example, the video clip with the puppy is nearly indistinguishable from a real video. The only thing that gives it away is the fact that the dog’s paw doesn’t compress when it makes contact with the ground. Also, with the video of the concert, the person’s hand on the left side is a little weird.

However, 99% of the people won’t be able to tell the difference. So, if those clips were put into a Hollywood movie or TV show, most people would think that they’re real. This also goes for the clip of the old man smiling or the marshmallows.

This is a pretty scary thing to think about, as we can tell that the CEOs of some large movie companies have their fingers on the “Fire Everyone and Pump Out AI Content” button. Let’s just hope that Google will exercise caution when distributing these models to the masses.