In partnership with

Your boss will think you’re a genius

If you’re optimizing for growth, you need ecomm tactics that actually work. Not mushy strategies.

Go-to-Millions is the ecommerce growth newsletter from Ari Murray, packed with tactical insights, smart creative, and marketing that drives revenue.

Every issue is built for operators: clear, punchy, and grounded in what’s working, from product strategy to paid media to conversion lifts.

Subscribe for free and get your next growth unlock delivered weekly.

Sign up now

OpenAI Just Dropped Sora 2, and Things Are About to Get Really Weird

Look, I've been watching the AI space long enough to know when something's actually a big deal versus when it's just Silicon Valley hype. And folks, Sora 2 is legitimately wild.

OpenAI launched this thing on September 30th, and within 48 hours it shot straight to #1 on the App Store—beating out their own ChatGPT app and Google's Gemini. That should tell you something right there.

What Makes This Different From Every Other AI Video Tool

Here's the thing: we've had AI video generators before. They were... fine? Mostly good for creating weird, melting nightmare fuel that looked vaguely like what you asked for. You'd type "person doing a backflip" and get someone whose spine bent in directions that would make an exorcist nervous.

Sora 2 actually understands physics. Like, real physics. It knows that basketballs bounce off backboards instead of teleporting through them. It knows that when you do a backflip on a paddleboard, the board should bob in the water realistically. It can generate Olympic-level gymnastics routines where bodies actually move like bodies should move.

That's the technical leap everyone's missing in the headlines. OpenAI essentially taught their model that reality has rules, and those rules matter. Previous models would just morph and distort the world to match your prompt. Sora 2 will show you failure states—because that's what actually happens in the real world.

But Wait, There's More: You Can Put Yourself In These Videos

Here's where it gets genuinely unsettling in the best/worst way possible.

The "Cameo" feature lets you create a digital version of yourself—or anyone else who gives permission—and drop that avatar into AI-generated scenarios. The process takes about a minute: you record yourself reading numbers aloud, move your head around so it can map your face, capture a voice sample, and boom. You've got a digital you.

Then you can make that digital you do... basically anything. Dance in Times Square. Give a TED talk. Apparently, people are already making videos of Sam Altman shoplifting, because of course they are.

And it's not just faces—OpenAI says it works with "any human, animal or object." So you can Cameo your dog, your car, your favorite coffee mug. The possibilities are both endless and slightly terrifying.

The Sound Changes Everything

People are sleeping on this, but Sora 2 generates synchronized audio. That means dialogue in multiple languages, environmental sounds that match the scene, sound effects that line up with actions, and—this is the kicker—lip movements that actually match the generated speech.

Previous AI video was essentially silent film era stuff. This is the jump to talkies. That's not incremental; that's transformative.

Imagine personalized video messages at scale. Imagine stock footage that doesn't exist but sounds like it was professionally recorded on location. Imagine never needing to hire voiceover actors for rough cuts again.

The Critical Impact: Three Things That Should Keep You Up at Night

1. The Death of "Seeing Is Believing"

We're about to cross a threshold where the average person cannot reliably distinguish between real footage and AI-generated content. Sure, OpenAI is adding watermarks and metadata. But let's be real—how many people check metadata? How many people even know what C2PA provenance data is?

The social contract around visual evidence is about to get shredded. Court cases, journalism, personal disputes—anything that relies on "here's video proof" is entering dangerous territory. We're not ready for this.

2. The Creator Economy Gets Disrupted (Again)

Stock footage libraries, voiceover artists, background actors, location scouts—these industries are staring down the barrel of automation. Why pay for stock footage of a sunset when you can generate an infinite number of custom sunsets for free?

The democratization argument goes: "Now everyone can be a filmmaker!" The economic reality is: "Now established filmmakers can fire half their crew." Both things can be true simultaneously, but one has more immediate consequences for people's rent payments.

3. Identity and Consent Are About to Get Complicated

Yes, OpenAI built in consent mechanisms for Cameos. You can revoke access, control who uses your likeness, delete videos. That's good! That's responsible!

But here's what worries me: the technology exists now. If OpenAI can do it with consent frameworks, someone else can—and will—do it without them. We're maybe 18 months away from anyone being able to generate convincing video of anyone else saying anything. The Cameo feature is essentially a proof of concept that this is now trivially possible.

The Bigger Picture: World Simulation

OpenAI keeps talking about Sora as a step toward "general-purpose world simulators" and training robotics. That's not marketing fluff—that's the actual endgame.

If you can build an AI that accurately models physics, understands how objects interact, and can predict outcomes in three-dimensional space, you've got the foundation for robot brains. You've got the foundation for training AI agents in simulated environments before they touch the real world.

Sora 2 isn't really about making cool videos for social media. It's about teaching AI systems to understand and predict physical reality. The video generation is almost a side effect—an incredibly lucrative, culture-shifting side effect, but still.

What This Means Right Now

OpenAI is rolling this out slowly: iOS only, US and Canada only, invitation codes that spread through Discord and social media. It's the classic "controlled explosion" strategy. They know this is powerful, potentially dangerous, and definitely going to cause chaos in multiple industries.

The free tier is surprisingly generous, which tells you they're optimizing for adoption over immediate revenue. They want this to become infrastructure, the way GPT did for text.

ChatGPT Pro subscribers get the full-powered version—20-second videos at 1080p instead of 10 seconds at 720p. The roadmap talks about 4K, real-time generation, multi-minute videos, interactive editing. Android support is coming Q4 2025 or Q1 2026.

The Uncomfortable Truth

Look, this technology is genuinely impressive. The physics modeling alone represents a legitimate breakthrough. The synchronized audio is legitimately cool. The Cameo feature is legitimately useful for a thousand creative applications.

But we're also hurtling toward a future where manufactured reality is indistinguishable from recorded reality, and we don't have the cultural antibodies for that yet. We don't have the legal frameworks. We barely have the vocabulary to discuss what this means.

OpenAI is doing more than most to build responsibly—the consent mechanisms, the moderation systems, the red team testing. But "more than most" might not be "enough" when you're fundamentally disrupting how humans perceive truth itself.

The Sam Altman shoplifting videos are funny now. They won't be funny when it's your face in a compromising position, and your employer doesn't understand how AI generation works.

Bottom Line

Sora 2 represents OpenAI moving from "impressive tech demo" to "actually useful tool that people will use every day." That's the GPT-3.5 moment they're talking about—when the technology crosses from "neat" to "essential."

Whether that's exciting or terrifying probably depends on whether you're in an industry about to get automated, whether you're thinking about the creative possibilities or the abuse vectors, and whether you're generally optimistic or pessimistic about humanity's ability to adapt to exponential technological change.

Me? I'm fascinated and worried in roughly equal measure.

The genie's out of the bottle now. We're all going to have to figure out how to live in a world where video can be conjured from imagination, where your face can be anywhere, and where "I saw it with my own eyes" no longer carries the weight it used to.

Welcome to the future. It's going to be really, really weird.