Only a handful of developers and artists have had the opportunity to utilize OpenAI’s Sora video-generation application, which created a buzz upon its initial release in February. Although OpenAI had initially indicated a full launch in the near future, recent updates suggest that the release may not be as extensive as anticipated. According to Mira Murati, the chief technology officer at the company, Sora could potentially be available to the general public within a few weeks.
The excitement surrounding the debut of Sora last month quickly dissipated, as it became apparent that Sora was not pioneering a new concept—numerous text-to-video tools are already on the market. However, what sets Sora apart is its ability to produce lifelike videos, whether it’s crafting realistic scenes, mimicking 3D animations, or creating cartoons. The quality of the output has garnered significant attention from the film industry, with reports suggesting that filmmaker Tyler Perry halted an $800 million studio expansion project after witnessing the capabilities of Sora.
To ensure a secure and controlled release, OpenAI is restricting access to trusted partners, positioning Sora’s launch as a carefully managed process. Rushing the public release of Sora is deemed unwise, given the potential misuse of AI technologies for malicious purposes, particularly in the midst of an anticipated volatile election year in the United States. Murati has hinted that the finalized version of Sora may not have the capability to replicate famous individuals.
One recent example shared by OpenAI with the WSJ depicted a princess reviewing a smartphone, showcasing the application’s capabilities. Credit: OpenAI
In an interview with the WSJ, Murati faced inquiries regarding the data used to train Sora. OpenAI, known for leveraging vast amounts of text data from across the internet to train its models, is currently embroiled in a legal dispute with The New York Times over copyright infringement claims related to text data. While Murati remained vague about the specifics, she mentioned that the training data consisted of “publicly available and licensed information,” with Shutterstock being one of the collaborating partners.
As OpenAI prepares for Sora’s launch, efforts are underway to address its resource-intensive nature. Sora’s operational costs exceed those of the company’s other AI applications, prompting initiatives to optimize expenses, particularly in comparison to the Dall-E electronic image generation tool. Failure to streamline Sora’s operational costs may lead to premium access charges. Furthermore, OpenAI is exploring the integration of audio capabilities into Sora videos, a development that would further escalate operational expenses.
