How to Use Sora: OpenAI's New Video Generation Tool Explained

Today, OpenAI Released Its Video Generation Model Sora to the Public

OpenAI has officially launched its new video generation model, Sora, on the fifth day of its "Shipmas" event, a 12-day marathon of tech releases and demos. Here's everything you need to know about the model—and how you can start using it today.

What is Sora?

Sora is an advanced AI-powered video generation model that can create videos from text prompts, animate images, or remix videos in unique styles. Though OpenAI first previewed Sora in February, today marks its public release for broader use.

What’s New About This Release?

While Sora’s core function—creating impressive videos from simple prompts—remains unchanged from the February preview, OpenAI has optimized the model to be faster and more cost-effective ahead of its public launch. The new features include:

• Storyboard: This feature lets you create multiple AI-generated videos and assemble them on a timeline, much like traditional video editing tools such as Adobe Premiere Pro.

• Creative Feed: The feed functions as a gallery where users can post their Sora-generated videos, see the prompts behind them, and make tweaks for inspiration.

How Much Can You Do with Sora?

Sora allows you to generate videos from text prompts, remix existing videos, and change their style or elements. The Remix tool lets you adjust different aspects of a video, while Storyboard enables you to assemble multiple clips together. Sora also offers preset styles, such as moody film noir or stop-motion effects like cardboard and papercraft. Additionally, you can trim and loop the videos you create.

Who Can Use Sora?

To generate videos with Sora, you need a subscription to one of OpenAI’s premium plans—either ChatGPT Plus ($20/month) or ChatGPT Pro ($200/month). Both plans provide access to other OpenAI products.

• ChatGPT Plus users can generate videos up to 5 seconds long, with a resolution of 720p, and create up to 50 videos per month.

• ChatGPT Pro users enjoy more flexibility, including generating videos up to 20 seconds long, with a resolution of 1080p, and creating up to 500 videos per month. Pro users can also generate up to five variations of a video per prompt and create unlimited "relaxed" videos, queued for generation during lower traffic times. Pro subscribers can create videos in vertical, horizontal, or square formats.

For those without a subscription, you can still view the feed of Sora-generated videos but won’t be able to create your own.

OpenAI is starting its global launch of Sora today, but it will take longer to launch in “most of Europe,” the company said. 

Where Can I Access Sora?

OpenAI has separated Sora from ChatGPT, and you can now access it directly at Sora.com by logging in with your ChatGPT Plus or Pro account. However, at the time of writing, we were unable to access the site due to heavy traffic, with a notice indicating that signups were temporarily paused.

How Did We Get Here?

Since OpenAI first introduced Sora in February, other tech companies have rolled out their own video generation tools, such as Meta Movie Gen and Google Veo. However, Sora’s release has also sparked controversy. Some artists, who had early access to experiment with the tool, leaked it in protest against OpenAI’s decision to train the model using artists' work without compensation.

What’s Next?

As with any new model, it’s unclear what measures OpenAI has put in place to prevent Sora from being used for malicious, illegal, or unethical purposes, such as creating deepfakes. On the topic of moderation and safety, an OpenAI employee acknowledged that they "might not get it perfect on day one."

Another pressing concern is the significant computing power and energy required to generate videos with Sora. Video creation consumes far more computational resources—and thus energy—than generating text responses in ChatGPT. The ongoing AI boom has already raised concerns about its environmental impact, and the widespread use of Sora and similar video generation models could exacerbate this issue.