OpenAI "Video AI 'Sora' Available to General Public in Second Half of Year"

Plan to Provide Integrated Audio and Video Editing Features

OpenAI's video generation artificial intelligence (AI) 'Sora' is expected to be commercialized in the second half of this year.

OpenAI CTO Mira Murati said in an interview with the Wall Street Journal (WSJ) on the 13th (local time), "The public will be able to use Sora this year," adding, "It could be in a few months." Although no specific timeline was mentioned, WSJ predicted an official launch in the second half of this year.

[Image source=AP Yonhap News]

Sora is an AI that generates videos based on user-input prompts. OpenAI introduced Sora as "capable of quickly producing videos up to one minute long containing multiple characters, specific types of actions, and complex scenes," and stated that "it can accurately interpret user prompts based on a deep understanding of language and generate vivid characters."

CTO Murati revealed plans to "integrate sound" to make the videos generated by Sora more realistic. She also mentioned that they are considering ways to allow users to freely edit video content created by Sora.

Regarding the training data for Sora, she said, "We use publicly authorized data as well as content from our partner, the image and video content company Shutterstock." She also explained, "Sora requires much higher operational costs than other AIs," and added, "We are working to reduce the cost to a level similar to OpenAI's image generation AI 'DALL·E.'"

Addressing concerns about the spread of AI-generated fake news in this global election year, CTO Murati stated, "When Sora is released, it will block the generation of images of public figures, similar to the policy for DALL·E," and added, "Videos will also display watermarks to distinguish them from real footage."

Currently, Sora is being operated with access limited to a small number of users since its initial release. Before commercializing Sora, OpenAI is conducting safety tests with a 'red team' composed of film producers, visual artists, and designers. The red team is a simulated adversary organized by the company to identify flaws and vulnerabilities in AI systems, contributing to strengthening the software's security.

This content was produced with the assistance of AI translation services.