It’s been nearly 24 hours since Google’s I/O 2025 keynote, and I’ve almost extricated myself from the deluge of announcements. I’m told there is a sky out there with breathable air, but I’ll report back on the veracity of those claims once I’ve seen it for myself. A lot transpired yesterday, but also, in a way, very little: there was AI for fanatics, AI for glasses, and a new AI with the potential to exacerbate your body dysmorphia. What I’m saying is there was one thing (AI), but an absolute s***load of it. While some of that AI is (theoretically) practical stuff, like a new mode for AI search or advanced research models for heavy number-crunching, that’s not the AI that caught my eye necessarily. For me, it was all about Flow.
Flow is a new product that Google is calling an “AI filmmaking tool,” and it combines Veo 3, the company’s latest video generation model, Imagen, its text-to-image model, and Gemini, Google’s all-purpose large language model, which powers its ChatGPT competitor. With that holy trinity, you can create fully generated videos from scratch—and not just the video part, either, but sound too. One of the most interesting aspects of Flow is that, thanks to text-to-audio abilities, it is actually pretty comprehensive. Here’s a cursed clip of a talking muffin to get a sense of what I mean.
My first Veo 3 gen
> a video with dialogue of two muffins while baking in an over, the first muffin says “I can’t believe this Veo 3 thing can do dialogue now!”, the second muffin says “AAAAH, a talking muffin!” pic.twitter.com/VA2VUZF8sS
— fofr (@fofrAI) May 20, 2025
And it’s not just audio that makes Flow unique, it’s the tool’s ability to extend and manipulate scenes through camera controls (choosing angles, motion, and perspectives) and continuity. “Once you’ve created a subject or a scene, you can integrate those same ingredients into different clips and scenes with consistency. Or you can use a scene image to start a new shot,” says Google. That means, in theory, you can really bite off something scene by scene and stitch those elements together for a longer final creation.
The thing about Flow is that, on one hand, it is kind of magical. In just a couple of years, we’ve gone from Will Smith eating spaghetti to being able to craft whole-ass animated movie scenes using just text-to-video/audio. However you feel about AI, that’s objectively an impressive technical feat. On the other hand, though, the philosophy behind tools like Flow is nothing short of bleak. In theory, being able to conjure up a movie using just some text inputs and a laptop is a watershed moment for movies, but in practice, I’m not sure it will play out that way. If I’m to let my pessimistic brain take the wheel here, I think it might lead to a lot more of something, and art is not that thing—I’m talking about AI slop.
As much as making visual art in the modern age has benefited from the infinitely accessible digital toolset we now enjoy, those tools (like Photoshop or Final Cut, for example) are just that: tools. They aren’t there to make anything for you, they’re there to take what you’ve made with your flesh computer (your big human brain with all of its fold-y goodness) and realize the vision. I’m here to say that’s a good thing. So much about art has nothing to do with form or style but about what it communicates from one human to another. A communication from me to you. And the problem with generating video like Flow from start to finish—from idea to creation to refinement—is that it eliminates that human element entirely. To be frank, that sucks, and I don’t know about you, but if I were capable of being entertained or enriched by a robot, I’d still be talking to Smarter Child on AIM.
To be fair, I think Google does see Flow as an augmentative tool and not a replacement, at least to an extent. It’s even partnered with some real filmmakers to show you that its AI is worthy of your serious creative endeavors. I don’t doubt that there will be some who find it genuinely useful, too, and what they make might even rise to the level of art. But mostly I predict Flow will be far more popular with those who would rather not bother with ideating or creating at all, and a filmmaker that type of person is not. I guess we’ll see one way or another soon enough since Flow is available via Google Labs right now. Call me a pessimist, or a Luddite, or utterly lacking imagination, but whatever you call me, just make sure you thought of it yourself. Trust me, life is more exciting that way.
