‘Sora’ is set for public availability later this year.
| Photo Credit: Reuters

OpenAI’s highly anticipated text-to-video generator, Sora, is set to debut later this year as revealed by OpenAI’s chief technology officer, Mira Murati, according to a report by The Wall Street Journal. Murati hinted that the release could occur within “a few months.”

Initially showcased in February, Sora boasts the ability to craft hyperrealistic scenes based solely on text prompts. Although initially accessible only to visual artists, designers, and filmmakers, Sora-generated videos have already surfaced on platforms like X.

OpenAI plans to expand Sora’s capabilities by eventually integrating audio, aiming to enhance scene realism further. Moreover, users will gain the ability to modify the content produced by Sora, acknowledging the occasional inaccuracies inherent in AI-generated imagery.

Despite questions about the data used to train Sora, Murati remained tight-lipped, only disclosing that it relied on publicly available or licensed data. She confirmed that Sora utilizes content from Shutterstock, with which OpenAI has a partnership.

(For top technology news of the day, subscribe to our tech newsletter Today’s Cache)

Murati acknowledged that powering Sora is “much more expensive” than previous AI models but assured that OpenAI is striving to maintain affordability. Similar to DALL-E, OpenAI’s text-to-image model, Sora will likely come with policies restricting the creation of images of public figures and will feature watermarks to distinguish its output from genuine content.