OpenAI launches Sora video generator for ChatGPT subscribers

OpenAI launches its anticipated AI video generator, Sora, on Sora.com today, but with limited availability.

The service will be available in the US and many other countries, but notably absent from most EU countries and the UK. While OpenAI CEO Sam Altman hasn't explained the regional restrictions, he indicated the wider rollout might take some time.

ChatGPT Plus and Pro subscribers can access Sora through their existing subscriptions at no extra charge, even though it's designed as a standalone product.

Plus users are limited to 50 video generations monthly, while Pro subscribers get a better deal: unlimited videos at standard processing speed, 500 fast-processed videos, and the ability to create higher-resolution content.

Sora pricing | Image: OpenAI

The version launching today is what OpenAI calls a "turbo" variant—a streamlined version of the Sora model introduced in February, optimized for cost and speed.

The system lets users create videos in several ways: by typing text descriptions, turning still images into videos, or modifying existing videos. It also includes features for remixing video content and more.

First review reveals impressive visuals, but physics still needs work

Tech reviewer Marques Brownlee (MKBHD) spent a week putting Sora through its paces and found both strengths and weaknesses. The AI shines at creating beautiful abstract scenes and cartoons, but stumbles when it comes to following basic laws of physics.

Objects in Sora's videos tend to appear and vanish at random or move in ways that defy physics, Brownlee found. This shows a key limitation of current AI: while it can match patterns to create convincing content, it doesn't actually understand how the physical world works—something OpenAI hopes to improve with future Sora research.

The physics problems show up most clearly when Sora tries to animate people and animals, with legs sometimes jumping into impossible positions. But it's not all bad news—Brownlee noted that the AI handles effects like flowing water and fire with remarkable accuracy.

Generation times vary by resolution: 480p videos take less than 20 seconds to create, while complex 1080p videos might need several minutes to complete.

According to Brownlee, Sora excels at creating abstract content, text animations, and cartoons. The cartoon style works especially well because any physical oddities can simply look like artistic choices.

While OpenAI has built in safety measures like watermarks, Brownlee worries about how the technology could be misused to create misleading content. He sees Sora as just the beginning of a future where AI-generated videos will be everywhere.

Note: This article is being updated with new information as it becomes available.