OpenAI has announced the latest version of its AI-powered text-to-video generator and the results are seriously impressive.
Announced on Tuesday, Sora 2 arrives alongside the all-new Sora social app that lets you “turn your ideas into videos and drop yourself into the action” — but more on that in a moment.
First, Sora 2. OpenAI’s first major update to its AI video generator since the original version’s release in February 2024 now offers audio — a huge improvement that puts it on equal footing with Google’s Veo 3 video generator.
OpenAI shared a selection of clips (below) created entirely by Sora 2. “Everything you are about to see and hear was generated by Sora 2,” the company said in a note accompanying the video.
The segments include everything from an ice skater with a cat on her head to a guy trying to do a backflip on a paddleboard. There’s also a dog astronaut, something strongly resembling Japanese anime, and a guy trying to ride two horses at once before it all goes horribly wrong.
Check out this stunning example, too:
The quality of the imagery is pretty amazing, with Sora 2’s AI able to render challenging content like water splashes with astonishing realism.
“Sora 2 can do things that are exceptionally difficult for prior video generation models,” OpenAI said. “It’s more physically accurate and realistic than prior systems and a big leap forward in controllability.”
OpenAI pointed out how its earlier video models can be what it described as “overoptimistic,” explaining that they would “morph objects and deform reality to successfully execute upon a text prompt. For example, if a basketball player misses a shot, the ball may spontaneously teleport to the hoop. In Sora 2, if a basketball player misses a shot, it will rebound off the backboard.” But not anymore.
It added that Sora 2, which for now is invite-only, is “far from perfect and makes plenty of mistakes,” though it’s continuing to work to make the videos even more real.
You can also do some really wacky stuff by directly injecting elements of the real world into Sora 2. For example, the AI observed a video of an OpenAI employee before Sora 2 gave him a trumpet and surrounded him with galloping zebras.
Indeed, it’s this “upload yourself” capability that inspired OpenAI to create the Sora social app for iOS.
Powered by Sora 2, the app lets you “create, remix each other’s generations, discover new videos in a customizable Sora feed, and bring yourself or your friends in via cameos” that let you drop yourself into any Sora scene.
The Sora iOS app is available to download now in the U.S. and Canada, with other countries coming soon. However, for now, you’ll need an invite to activate the app.
The new and improved Sora will no doubt send another shockwave through the creative industries as artists try to get their head around the full impact of the technology on their work.
Digital Trends recently tested three AI video generators with the same prompt to see which gave the best result. Find out which came out on top.