Google I/O was about a week ago, and if you haven’t heard, one of Google’s biggest announcements was the company’s Veo 3 generative AI model for video. Gone are the days of creepy, low-quality clips that vaguely look like Will Smith eating spaghetti and don’t traverse the uncanny valley very well. Veo 3 is more than capable of generating that with near perfect photorealism, and it can do a lot more than that, too.
“Seed oils made y’all weak,” this AI-generated phantom says.
The key upgrades in Veo 3 aren’t related to simple photorealism; OpenAI’s Sora was doing that last year. Instead, Veo 3 supports the ability to have multiple unique and consistent characters with dynamic character movements and native audio. That means you get natural-sounding dialogue—even with accents—sound effects, and even music all baked right into the model. No more having to produce (or generate) separate audio tracks for your AI video; Veo 3 does it all in one, and it’s shockingly good at it.
Text on the cover of this man’s book remains consistent throughout, even if some of it is illegible.
If you want proof, all the images in this post are from a video produced by an “AI Video Production company” out of Germany known as The Dor Brothers. The video, titled Influenders, parodies many different types of video “influencer” by representing how they would behave during the apocalypse in short clips. It was made with Google’s Veo 3, and this format is ideal for the new tool because it is actually limited to short, eight-second clips.
Most of the text in this parody of a Pride influencer is illegible.
The video is incredibly impressive, and could easily fool many people who aren’t watching closely. There are plenty of AI artifacts to be found, still, but you really do have to actually be looking for them, at least in most of the scenes. The video and sounds are shockingly realistic, even including the background noises of crowd panic and end-of-the-world destruction.
Characters in the background of this shot warp, but it’s not obvious unless you’re looking for it.
Though Veo 3 is limited to eight-second clips currently, Google has already created a tool called Flow that allows the model to retain its context between generations. That means that once you’ve created a scene or character, you can integrate those same components into new generations and clips. This, along with careful video editing, can allow for the creation of longer videos without needing the ‘cut every eight seconds’ structure of Influenders. If you’re keen on trying out Veo 3 and Flow for yourself, you’ll need to be a subscriber to Google’s AI Ultra or AI Pro plans.
You can check out Influenders above; it’s worth a watch just so you can laugh at the parodies of the most irritating people on the Internet. Everyone gets hit; it’s not apolitical, but it’s pretty even-handed in its lambasting. After you’ve watched it, you can join the author in having a minor existential crisis. Are we real? Or are we just prompts?
