OpenAI, the pioneering artificial intelligence research laboratory, is gearing up to unveil its highly anticipated text-to-video generator, Sora, to the public later this year. In an exclusive interview with The Wall Street Journal, Mira Murati, OpenAI's Chief Technology Officer, disclosed that Sora's release is imminent, possibly within the next few months.
Initially showcased in February, Sora has garnered considerable attention for its ability to create hyperrealistic video scenes based solely on textual prompts. While initially restricted to visual artists, designers, and filmmakers, some Sora-generated videos have already surfaced on platforms like X, showcasing the tool's remarkable capabilities.
Murati revealed that OpenAI has ambitious plans for Sora's future development. The company aims to integrate audio functionality, enhancing the realism of the generated scenes. Moreover, users will have the opportunity to edit the content produced by Sora, addressing concerns about the accuracy of AI-generated imagery. Murati emphasized, "We're trying to figure out how to use this technology as a tool that people can edit and create with."
Despite queries regarding the data used to train Sora, Murati remained tight-lipped, stating only that it comprised publicly available or licensed data. While she did confirm a partnership with Shutterstock, she refrained from elaborating on the inclusion of content from platforms like YouTube, Facebook, and Instagram.
One significant aspect highlighted by Murati was the cost associated with powering Sora. She admitted that Sora is "much more expensive" to operate but assured that OpenAI is striving to maintain affordability, aiming for costs comparable to DALL-E, the company's existing text-to-image model.
As the release of Sora looms closer, concerns about the potential misuse of generative AI tools, particularly in spreading misinformation, have intensified. Murati reassured that Sora's capabilities would likely be restricted from generating images of public figures, aligning with DALL-E's policies. Additionally, videos produced by Sora will bear watermarks to distinguish them from authentic content, although concerns remain about the effectiveness of such measures.