Last Thursday, OpenAI switched off its Sora video API. Within about a day, two of the people behind The Lion King were talking publicly about making films with AI, and Kling announced a model that roughly doubles how long a single shot can run. None of this is finished work yet, and I’ll be clear about that as we go. Still, I find the direction fascinating, so let’s walk through what happened, what the skeptics are saying, and what I think you can do about it right now.
Sora goes quiet
OpenAI’s video generation guide now opens with a simple note: “The Sora 2 models and Videos API were shut down on September 24, 2026 and are no longer available. No one-to-one replacement API is available.” The deprecations page shows that developers were told back on March 24, and the replacement column for every Sora model is just a blank dash. If you built on Sora, you had six months of warning, and now you’ve got a migration on your hands.
Katzenberg wants to show what’s possible
On September 25, Jeffrey Katzenberg told TheWrap’s Sharon Waxman that he’s creating “either a digital platform or an actual studio” to make animated film and television with AI. “I am fired up, and going to go out and do something that actually shows what’s possible here,” he said. He was also candid about how he felt until recently: “Up until recently it was slop.”
When Waxman asked what his biggest barrier would be, he answered with one word, talent, and added, “That’s the hard part.” I love that answer, because it puts people at the centre. There’s no company name or timeline yet. Earlier in September, The Verge, citing The Information, reported that Katzenberg was teaming up with former Sora lead Bill Peebles on an unnamed video startup. I honestly can’t tell from the coverage whether that’s the same effort or a separate one.
A Lion King director signs on for Storm Dogs
The same day, Variety reported that Rob Minkoff, co-director of The Lion King and director of Stuart Little, is in advanced discussions to be creative lead on Storm Dogs. It’s a family film from Particle 6, the company behind the AI performer Tilly Norwood, about a toddler and a pack of dogs finding their way home after a storm. Particle 6 says it will combine “live-action performances with photorealistic animated characters,” and production is targeting a 2027 start. Particle 6’s first feature, Misaligned, began production earlier this month.
What struck me is how Particle 6 describes the work. Particle 6’s Eline van der Velden talks about “keeping every creative decision in human hands,” and Minkoff talks about animal characters who are “funny, emotionally truthful and completely alive.” That’s a storyteller’s framing, and I think it matters.
Kling stretches the shot
Then on September 28, Kling announced Kling 4.0 for October, with 4.0 Flash live now for Ultra Yearly subscribers. According to CellCog’s write-up of the announcement, the headline features are native 30-second generation (Kling 3.0 does 3 to 15 seconds), up to 10 keyframes, up to 15 multimodal references, 4K with 10-bit HDR, stereo audio, and more accurate lip sync. Please treat those as claims. They’re reported secondhand, Kling hasn’t published benchmarks, and I haven’t seen anyone test them independently.
On the research side, Google Research has shown an AI video co-director built from agentic frameworks on top of Gemini and Veo, and one of them produced a continuous 10-minute film. Keeping characters and props consistent across minutes is exactly the problem long-form storytelling needs solved.
The case for staying skeptical
There are good reasons to hold all of this lightly. Everything above is intent and development. Katzenberg has no named company, no announced funding and no slate, and TheWrap itself brings up Quibi, which raised $1.75 billion and folded within six months. Minkoff’s deal isn’t closed, Storm Dogs won’t start production before 2027, and Kling’s numbers are untested.
You can also argue that Sora’s exit was a formality. Plenty of people would tell you studios and creators had already moved on to Veo, Kling and Runway long before the lights went out. And the creative community is far from united. In Austin, the Austin AI Film Festival runs October 16 to 17 at Zach Theatre and has drawn real backlash, while the separate Austin Film Festival doesn’t accept films that use AI in their writing or creation and requires disclosure of any AI use. Those concerns deserve respect, and I don’t think they’ll go away.
What you can do now
Here’s where I land. When experienced storytellers start saying the tools are good enough to be worth their time, that’s a signal worth paying attention to, even before the films exist. And you don’t need to wait for a studio to get started.
If you’re a developer who built on Sora, look at your pipeline and put an abstraction between your app and any single video model. Providers come and go, and the Sora shutdown is a good reminder to design for swapping engines. If you’re curious about long-form work, the Google co-director code for two of the frameworks is already on GitHub, so you can dig into how memory and planning keep a story coherent.
If you’re a creator, pick one short idea and make it end to end. Write the story first, storyboard it, and then use whichever tool fits each shot. Keep notes on where the models help and where they fight you. Be open about how you used AI, too, because festivals and audiences are asking, and honesty builds trust.
Talent was the barrier Katzenberg named, and that’s the part you already bring. I’d love to see what you make.