Ten months after its preview, OpenAI has officially released a Sora video model called Sora Turbo. Described as “hyperrealistic,” Sora Turbo generates clips of 10 to 20 seconds from text or image inputs. It outputs video in widescreen, vertical or square aspect ratios at resolutions from 480p to 1080p. The new product is being made available to ChatGPT Plus and Pro subscribers ($20 and $200 per month, respectively) but is not yet included with ChatGPT Team, Enterprise, or Edu plans, or available to minors. The company explains that Sora videos contain C2PA⁠ metadata indicating that they were generated by AI.

While Sora Turbo is being offered in the U.S. and numerous other countries, OpenAI has chosen not to release it at this time in the UK, Switzerland and the European Economic Area, pushing back against regulatory constraints.

The tool has physical limitations as well. OpenAI has chosen not to release it at this time in the UK, Switzerland and the European Economic Area.“Likeness is currently only available as a pilot feature to a small group of early testers,” and there are built-in safeguards, including algorithmic suppression of child abuse and scenes of violence or nudity, OpenAI writes in an announcement.

Available at Sora.com, the initial commercial version “often generates unrealistic physics and struggles with complex actions over long durations,” OpenAI notes, adding that “although Sora Turbo is much faster than the February preview,” the compute time and power necessary to generate video still make such services difficult to scale.

Plus subscribers can generate “up to 50 videos at 480p resolution or fewer videos at 720p each month,” while the Pro plan includes “10x more usage, higher resolutions, and longer durations.” The company says it is exploring additional price options to be announced early next year.

Marques Brownlee, a tech-loving YouTuber known as MKHBD, posted a Sora review demonstrating its capabilities in various scenarios. Brownlee, VentureBeat reports, “shared that while Sora could produce impressive and sometimes eerily realistic footage,” it also exhibits a tendency “to hallucinate random details,” garble display text and flub the laws of physics.

Ars Technica observes that since OpenAI first previewed the generative tool in February the market has been booming, with “video-synthesis models from competitors — such as Google’s Veo, Runway’s Gen-3 Alpha” as well as Kuaishou Technology’s Kling and MiniMax’s Hailuo AI “taking some of the shine off of Sora’s release.”

VentureBeat mentions Tencent, which surfaced Hunyuan Video this month, and cites Luma AI’s Dream Machine as a contender in “an increasingly competitive landscape for realistic, live-action AI video generation.”

“Still, it’s a major milestone for OpenAI to finally ship its highly anticipated video model,” notes Ars Technica, which embeds examples including a bubble-breathing dragon made of suds.