OpenAI has officially launched Sora, its text-to-video AI model, as part of its “12 Days of OpenAI” event. Sora enables users to generate videos from text prompts, animate images, and remix existing videos. ChatGPT Plus subscribers ($20/month) can create up to 50 priority videos at 720p, while ChatGPT Pro subscribers ($200/month) have access to unlimited video generations at 1080p. Do not be surprised if you encounter the “Sora account creation is temporarily unavailable” error message. Demand is understandably high.
To promote responsible use, all Sora-generated videos include watermarks and C2PA metadata indicating AI generation. Users must agree not to upload prohibited content. Currently, Sora is accessible in multiple countries, though its European release is delayed due to regulatory considerations. The launch follows a recent controversy where artists leaked the product in protest, accusing OpenAI of exploiting their unpaid labor during Sora’s development. This fight will continue for a very long time.
How good is Sora? It will never be worse than it is today – and today, it’s pretty good. It excels at videos where the laws of physics are suspended. In other words, animations work well because you accept that animated characters live in a universe where our laws of physics don’t apply. (Imagine a Road Runner cartoon where Wile E. Coyote runs off a cliff and continues for five steps as if the ground was still beneath him, then hangs in the air for a few seconds before falling hundreds of feet to create a Wile E. Coyote-shaped hole from which he emerges unharmed. Fun to watch, but it can’t actually happen in our universe.)
When you ask Sora to create an imaginary cartoon world, it’s very good. On the other hand, photorealism and cinematic generation still have a way to go. Things appear and disappear for no reason. Running and legs seem to pose a problem. There are lots of artifacts and other obvious unnatural movements that you are hardwired to see, so you instantly know something isn’t quite right. But, again, this is the worst Sora will ever be.
This is a launch that will be as profound to human communication as the advent of the mobile phone camera, social media, and social video; you can call it “social production.” In more than a few months (but less than a few years), anyone who can describe a video they want to see will be able to see it (and share it). Buckle up: the world will never be the same.
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.