Vallée Duhamel & Sora

OpenAI's latest blog post skips the usual spec sheet and hands the mic to two working filmmakers, Vallée Duhamel, who've been using Sora not as a toy but as a tool inside their creative process. The pitch is simple: instead of storyboarding a scene for weeks or waiting on a location scout, they type a description and watch a rough version of the world appear on screen within minutes.

What's interesting is how the duo frame Sora less as a replacement for cinematography and more as a sketchpad. They talk about testing atmosphere, lighting mood, and camera movement before a single crew member shows up on set. That's the part traditional pre-production tools never solved well. Storyboards are static. Animatics take time and money. Sora, in their telling, compresses that guesswork into something closer to daydreaming out loud.

There's also a subtler point buried in their comments about iteration speed. When generating a shot costs seconds instead of days, filmmakers can afford to be wrong more often. They can throw out ten ideas to find the one that works, rather than committing early because the alternative is too expensive. Vallée Duhamel describe this as expanding the range of what they're willing to try, not just speeding up what they already planned to do.

OpenAI, naturally, uses the story to argue that Sora is maturing from a flashy demo generator into something closer to a professional tool. That's a self-serving frame, sure, but the underlying behavior described, using AI video as a fast concepting layer rather than a finished-product machine, tracks with how other studios have quietly started experimenting with generative tools over the past year. Nobody's shipping a Sora-rendered feature film. But plenty of people are using it to figure out what the film should look like before spending real money finding out the hard way.