Close Menu
Saturday, April 19 Login
Art Generation

OpenAI’s Sora video generator appears to have been leaked

By No Comments6 Mins Read

A group appears to have leaked access to Sora, OpenAI’s video generator, in protest of duplicity on the part of OpenAI and what it calls “artwashing.”

The group published the project on Tuesday on Hugging Face, an AI development platform that appears to be connected to OpenAI’s Sora API, which is not yet publicly available. This group created a frontend that allows users to generate videos with Sora, presumably using authentication tokens from an early access system.

Through this group’s frontend, users can generate 10-second videos at up to 1080p resolution by entering a short text description. When TechCrunch tried it, the queue was quite long, but several users on X managed to upload samples, most of which had OpenAI’s distinctive visual watermark.

As of 12:01 PM ET, the front end has stopped working. I’d like to assume that OpenAI and/or Hugging Face revoked access.

The group claims that three hours later, OpenAI temporarily suspended early access to Sora for all artists.

So why did the group do this? OpenAI is pressuring Sora’s early testers, including the red team and creative partners, to weave a positive story around Sora, and they It alleges that the company has failed to compensate them fairly for their work.

“Hundreds of artists provide unpaid labor through bug testing, feedback, and experimental work on the (Sora Early Access) program of a $150 billion (sic) company,” says the company, which calls itself “Sora PR Puppets.” This group, which I will call , is attached to the frontend I wrote about in the post. “This early access program seems to be more focused on PR and advertising than creative expression and criticism.”

The group initially did not identify its members. But as the day progressed, several listings appeared in the Hug Face attachment and in a separate petition.

The group also claims that OpenAI is misleading early access users about Sora’s capabilities by locking them in tightly. The organization says all Sora output must be approved by OpenAI before it can be shared widely, and only a few creators in the program will be selected to screen their Sora creations.

“We are not opposed to using AI technology as a tool in art (if we were, we probably would not have been invited to this program),” the group wrote. “What we disagree with is how this artist program is being rolled out and how this tool is being prepared prior to general release. We believe that OpenAI is more open and artist-friendly. We are sharing this with the world in the hope that we can go beyond PR and support the arts.”

An OpenAI spokesperson said in a statement that Sora is still in “research preview” and the company is “committed to balancing creativity with robust safety measures for broader applications.”

“Hundreds of alpha artists shaped Sora’s development and helped prioritize new features and safety measures,” a spokesperson said. “Participation is voluntary and there is no obligation to provide feedback or use the tools. We are excited to be able to provide free access to these artists and through grants, events, and other programs. We believe that AI can be a powerful creative tool and are committed to making Sora useful and safe.”

The spokesperson also said artists have “no obligation” to OpenAI other than to use Sora “responsibly” and refrain from sharing sensitive details during Sora’s development. Ta. However, he did not specify what “responsible” use entails, nor did he say what details OpenAI considers sensitive.

Since its debut earlier this year, Sora has suffered technical setbacks as rivals in the video generation space work hard to overtake it. Not that it mattered, one of Sora’s co-leaders, Tim Brooks, left OpenAI for Google in early October.

In a recent Reddit AMA, OpenAI Chief Product Officer Kevin Weil said Sora is being held back by the “need to perfect the model, get things like security and impersonation right, and scale the compute.” I did. According to The Information, the original system released in February took more than 10 minutes of processing time to create a one-minute video clip.

Consistency was also an issue in early iterations of Sora. Filmmaker Patrick Cederberg had to generate hundreds of clips before he had a usable one, as the models struggled to maintain their style, objects, and character throughout the video.

According to code discovered by X users, the leaked Sora appears to be a faster “turbo” variant. The code also suggests style controls and limited customization options.

According to The Information, OpenAI has trained Sora on millions of hours of high-quality clips covering a variety of styles and subject matter to improve the quality of the videos it produces.

Technology-related hurdles aside, OpenAI has ceded valuable partnership ground to video generation challengers in recent months. In September, Runway signed a deal with Lionsgate, the studio behind “John Wick,” to train custom video models based on Lionsgate’s movie catalog. About a week later, Stability, which has developed its own set of video generation models, added Avatar director James Cameron to its board of directors.

OpenAI was said to be meeting with filmmakers and Hollywood studios to demo Sora earlier this year. Former CTO Mira Murati attended Cannes. However, the company has not yet announced any collaborations with major production companies.

Update: Added statement from OpenAI.

author avatar
See Full Bio
Share.

Related Posts

Add A Comment
Leave A Reply