The Camera Is Loaded and It’s Pointed at All of Us
What new SeeDance 2.0 video model means for filmmakers and what I just made on my laptop for next to nothing.
When I’m not directing documentary, films I help out with a small independent theater in Enterprise, Oregon called the OK Theater. It’s a historic building with a western flare that shows family films. Like all theaters, we needed a pre-roll telling people to silence their phones before the film starts.
A year ago, that would have meant either buying a generic video or producing something real. Even a short video would still mean locations, casting, crew, scheduling, travel, sound design, post. A modest production would take a week of planning, a week of shooting, and a week of post. Maybe $50,000 if you’re being reasonable about it. More if you’re not.
Instead, I used Google’s Nanobanana to generate reference frames for set and characters and VEO 3 through the Flow interface to animate eight second clips. Then I pulled everything into Adobe Premiere for editing and sound design. It took about a day. The result cost me almost nothing.
SeeDance 2.0
A new Chinese model called SeeDance 2.0 just dropped and it appears to have launched with essentially no intellectual property guardrails. Whether that was an oversight or a deliberate provocation is an interesting question. Either way, a video of Brad Pitt fighting Tom Cruise circulated across the industry and the reaction was not subtle.
Two things jumped out immediately. First, the action is good, really good. It prives that AI is over the “uncanny valley.” We’re not just talking about slow motion panavision of big breasted women lounging about in beautiful settings anymore. Second, and this is the one that makes the model different from others; it generates multiple camera angles.
Let me say it again. We’ve moved from 8-second single shots to 15-second multi-angle sequences, and that’s not an incremental improvement. The major issue with VEO 3, Sora and the others is that there is not consistancy between clips. The problem is solved, perhaps. The model seems to understand that cinema isn’t a moment, it’s a construction. Right now, the executives at Netflix are staring into mirrors wondering whether streaming companies like thiers are about to go into the same trash can they just spent a decade filling with the old studios.
The Gun Is Pointed at Me Too, and I Find It Strangely Interesting
I want to be honest about something uncomfortable. As a director, looking at what’s coming is genuinely terrifying. The industrial infrastructure of filmmaking; the crews, the stages, the development pipelines, the whole apparatus, is facing potential obsolescence, and the jobs that disappear probably aren’t coming back. I don’t think we are going to be retained as coders either, because Anthropic has destroyed that field already.
But I’d be lying if I said there wasn’t something tantalizing underneath the fear.
I have projects in my head that will never get made in the old system. Not because they aren’t good, but because I don’t have access to $100 million budgets, A-list talent, and the distribution relationships required to get something of that scale greenlit. Hollywood’s gatekeeping infrastructure filtered out most of what could have been made, not because it was bad, but because the economics demanded conservative bets.
Now I’m imagining assembling a small, elite team. Perhaps a writer, a production designer, an AI specialist, an editor, and a director of photography. to use next-generation models that can actually manifest those ideas. For science fiction especially, which I have a deep affection for, this could produce work that surpasses what was possible in the old world. Not despite the technology but because of it. The constraint was never imagination, it was set construction and special effects.
The problem for a legacy director, of course, is that this is available to everyone. My elite team will be competing against millions of films rather than thousands. It might be humbling. I tell myself that I have real advantages from my past experience creating story. The reality the biggest advantage is one that I don’t have much off, the ability to promote content on social media. Likely, that is more important than quality.
This competive window might last a year or two. Then AI will probably eliminate the need for any human at all. The logical endpoint is that someone on a couch at home could prompt a model for a personalized two-hour feature film, generated on demand, as technically accomplished as anything Hollywood ever produced.
It’s hard to imagine making a living in that world, when the value of entertainment is so low.
What Happens to Documentary?
This is where it gets personal, because documentary is my true love and my life’s work.
Here’s what I keep coming back to: documentary, by definition, is about the real world. The authority of the form comes from the fact that something actually happened, and we were there. That’s not replicable by a generative model, at least not in any honest sense.
So I suspect we’re about to see three categories where there used to be two. Human-made fiction will exist as a niche — smaller, more personal, valued precisely because humans made it, the way handmade furniture still has a market. Documentary survives as its own niche for the same reason authenticity survives: because reality has a texture that fabrication can’t fully counterfeit, and audiences eventually learn to feel the difference. There might be technology that vets there forms of film.
And then there’s the third category, which will dwarf both of them: AI-generated premium content, produced at volumes and price points that make the Netflix library look quaint. That’s the juggernaut. That’s where the money goes, where the casual viewer goes, where the algorithm points.
The question for those of us who’ve spent our lives making real films about real things is whether the niche is big enough to live in.
Cody is the founder of Rhumbline Media and the director of The Last Dive, winner of Best Cinematography at Tribeca 2025.