What is Sora 2, the AI model some say could be the end of Hollywood?

You know AI video generation is getting convincing when people start to suspect that genuine filmed footage is AI. Some of the posts on social media claiming Sora 2 to be the end of Hollywood are actually showing a Chat GPT ad campaign that was filmed on 35mm film.

Sora 2 does seem to be a big step forward in resolving some of the technical problems associated with AI video, producing clips with much more realistic physics and body mechanics. It's not going to replace camera operators yet, but OpenAI is now out to take on the likes of Instagram and TikTok using the model for its own Sora social media app, which features 100% AI-generated video and allows you to deepfake your friends.

What is Sora 2?

OpenAI says Sora 2 can do things that have been difficult if not impossible for AI video generation models until now. Its output is more physically accurate and realistic, more controllable, and it can generate synchronized dialogue and sound effects.

The original Sora was released in February 2024. Since then, the Sora team has focused on training models with more advanced world simulation capabilities, pre-training and post-training on large-scale video data. Example footage shows Sora 2 can generate video of things like gymnastics routines, which used to be horrific with previous AI models.

OpenAI says one problem is that video models are overoptimistic: they will morph objects and deform reality to successfully execute a text prompt. For example, if a basketball player missed a shot, the ball might teleport to the hoop. It says Sora 2 is much better at avoiding this and respecting the laws of physics.

The new model can also follow more intricate instructions spanning multiple shots while persisting world state. It can also insert real-world elements from reference footage, including human and animal subjects as well as objects.

Comparisons between Sora and Sora 2 emerging on social media give an idea of how much of an advance the new model has made. Users have also been testing the new model on anime styles.

A lot of early users are impressed, but predictions of the end of traditional filmmaking are premature. Sora 2 can generate longer clips than its predecessor, but it's still capped at 60 seconds. Getting consistency over multiple clips to assemble a coherent movie is still likely to be practically impossible.

Some say Sora 2 is regurgitating rather than generating some footage, with some generations appearing to suggest that it was trained on videos from YouTube.

"It spits back missions like a parrot with RTX. That’s not intelligence, that’s compression. trained on YouTube videos, clearly including several gameplay runs of this mission," one person wrote on X about its apparent ability to reconstruct an entire Cyberpunk side mission.

Where we probably will see Sora 2 used is for short social videos and ad campaigns – although perhaps not for ChatGPT. A lot of people are suggesting that the video below was generated by Sora. It was created by Isle of Any and shot on 35mm film by SMUGGLER director Miles Jay.

OpenAI is pushing its Sora iOS app⁠ for iOS as the way to access the new model. Once people have received an invite, they'll also be able to access Sora 2 through sora.com⁠.

Sora 2 will initially be available for free to get people to try it. ChatGPT Pro users will also be able to use the higher quality Sora 2 Pro model. There will also be an API.

If you want to make movies the traditional way, see our guides to the best video editing software. We've also reviewed the best video editing computers and the best laptops for video editing.