OpenAI's GPT Image 2 launched globally today, producing photorealistic 4K video from text prompts in under a minute , and the market is already repricing what professional video production is worth.
At 09:00 UTC this morning, OpenAI flipped the switch on what may be the most consequential product release in the company's history. GPT Image 2, built on the Omni-Attention architecture and internally designated Sora-v2, is now live for enterprise subscribers and Pro users worldwide. Within six hours of the API going live, more than 500,000 videos had been generated, briefly overwhelming OpenAI's status page in a demand surge that speaks for itself.
Sam Altman walked through a live demonstration on X, showing the model generating 4K clips up to 60 seconds in length , complete with procedurally matched audio and dialogue , directly from text prompts. The standard tier returns a finished 30-second clip in roughly 45 seconds. That's not a party trick. That's a production pipeline.
The 2024 Sora preview generated real excitement and real skepticism in equal measure. Anyone who used it seriously ran into the same wall: morphing hands, objects that defied gravity, scenes that couldn't hold visual continuity across more than a few frames. GPT Image 2's Omni-Attention architecture reportedly cuts those physical artifacting issues by 98% against prior industry benchmarks. That single number is why this release feels categorically different from what came before it.
The closed beta that ran through select Hollywood studios in late 2025 wasn't just a PR move. It gave OpenAI months of real-world feedback from professionals who work in physical continuity for a living , the people most likely to find the seams. That the product shipped broadly today suggests those sessions went well enough.
The market read the room immediately
Visual effects houses and stock media platforms took the sharpest hit, with some firms shedding up to 18% in pre-market trading. That's not a kneejerk reaction , it's investors doing the math on what happens to licensing revenue when a Pro subscriber can generate broadcast-quality footage from a sentence. Stock footage libraries in particular have a structural problem here: their value proposition has always been convenience and cost relative to a production shoot. GPT Image 2 undercuts both arguments simultaneously.
Semiconductor companies aligned with OpenAI moved in the opposite direction, a pattern that's become familiar as AI infrastructure demand keeps compressing the distance between model release and hardware scarcity.
Where this leaves the industry
The so-called Video Wars of late 2025 , Google's Veo, Runway's Gen-3 Alpha, and a half-dozen smaller contenders fighting for positioning , now look like a prologue. Today's release is the chapter where generative video stops being a research demonstration and becomes production infrastructure. Pre-visualization for film, lower-budget advertising, social content at scale: these are workflows where traditional CGI pipelines are about to face a genuine cost and speed argument they haven't encountered before.
That doesn't mean cinematographers and VFX artists disappear next quarter. The craft of directing, the judgment calls that shape a visual story, the ability to manage a client through a creative process , none of that gets automated by a text prompt. But the bottom of the market, the work that has historically funded studios while they pursue more ambitious projects, is under real pressure starting today.
Watch how the major advertising holding companies respond over the next 30 days. Their procurement decisions , whether they move budgets toward GPT Image 2 API access or hold the line on traditional production relationships , will tell us more about this technology's near-term commercial trajectory than any benchmark chart. The model is out. The negotiation over what it replaces has just begun.
Also read: OpenAI is preparing to launch Project Orion on May 14 and the AI industry is bracing for impact • A visual ghosting bug in OpenAI's latest image model is contaminating new generations with fragments from earlier in the same chat • GPT Image 2 just produced a fake video game screenshot that nobody can tell is fake
