OpenAI has announced that its artificial intelligence video generator, Sora, is now available to everyone in the U.S. The tool, which was initially introduced in February and limited to select artists, filmmakers, and safety testers, can now be accessed by the public. However, on Monday, OpenAI’s website faced heavy traffic and did not allow new sign-ups for Sora at various points.
Sora is a text-to-video generator, allowing users to create AI-generated video clips from written prompts. For example, one prompt on the website reads, “a wide, serene shot of a family of woolly mammoths in an open desert,” and the resulting video shows the creatures walking through sand dunes.
OpenAI hopes that this early version of Sora will inspire people to explore new forms of creativity, tell their stories, and push the boundaries of video storytelling, according to a blog post from the company.
OpenAI, known for its popular ChatGPT, has been expanding its work into various types of generative AI. The company is developing a voice-cloning tool and has integrated its image generation model, Dall-E, into ChatGPT. Backed by Microsoft, OpenAI is a leader in the growing AI market and is now valued at almost $160 billion.
Ahead of the public release of Sora, OpenAI allowed tech reviewer Marques Brownlee to test the tool. Brownlee described the results as “horrifying and inspiring at the same time.” He noted that while Sora excelled at creating landscapes and artistic effects, it struggled with accurately depicting basic physics. Some filmmakers who also tested the tool reported that it produced strange visual glitches.
OpenAI is currently addressing compliance issues related to the Online Safety Act in the UK, as well as the Digital Services Act and GDPR in the EU.
Two weeks ago, OpenAI temporarily suspended access to its Sora tool after a group of artists found a way to bypass restrictions and allow anyone to use it. The artists, calling themselves the “Sora PR Puppets,” accused OpenAI of “art washing” — a strategy they claimed would exploit artists and threaten their livelihoods. They argued that OpenAI was attempting to create a positive image for its product by associating it with creative individuals.
Although generative AI has made significant progress over the past year, it still struggles with issues like producing inaccurate responses, known as hallucinations, and plagiarism. AI image generators also frequently create unrealistic images, such as people with multiple arms or distorted facial features.
Critics caution that AI video technology could be exploited by malicious actors for purposes like disinformation, scams, and deepfakes. There have already been deepfake videos circulating, including one featuring Ukrainian President Volodymyr Zelenskyy allegedly calling for a ceasefire, and another with Vice President Kamala Harris supposedly referring to herself as “the ultimate diversity hire.”
OpenAI stated in its blog post that it would initially restrict uploads featuring specific individuals and block content containing nudity. The company also mentioned that it would prevent particularly harmful forms of abuse, such as child sexual abuse material and sexual deepfakes.
Sora will be accessible to users who already subscribe to and pay for OpenAI’s tools. While the tool will be available in the US and most other countries worldwide, it will not be offered in the UK or Europe due to copyright concerns.