Filmmaker: OpenAI's Sora doesn't mean Hollywood is doomed

Hollywood hasn’t yet gone into a panic over OpenAI’s recently announced impressive text-to-video generator.

Last week, OpenAI released a text-to-video generator called Sora, sparking a new round of discussion about how AI can change the entertainment industry. The model can quickly generate relatively high-quality videos from text, which has attracted a lot of attention.

However, industry insiders believe that Sora and other text, image and audio generation models still have some flaws, and entertainment industry creators need to balance concerns about their jobs and the limitations of AI capabilities.

“A lot of people say Hollywood is done, but I don’t think so,” said Reid Southen, a film concept artist and illustrator. “The production process is too complicated, and there are still many problems with these videos, especially defects such as time consistency and extra limbs.”

Concerns about generative AI came to the forefront for those working in the entertainment industry during a labor strike last year. Many creatives continue to push for protections to prevent AI models from being misused. Aubry Mintz, interim executive director of the Hollywood chapter of the International Animated Film Association, said the animation community is already "quite uneasy" about Sora.

Mintz believes that if Sora makes roles like reference animation, concept art, and storyboards redundant, these industry professionals could lose their jobs.

Southen doesn't think Sora will replace filmmakers anytime soon, but he worries about the impact of advanced technology on future jobs in the industry. "We obviously live in a 'just enough' culture, so if it saves money, people and companies will do it," he said.

Blake Ridder, a British director, writer and actor, said that at present, Sora's capabilities seem more suitable for creating stock footage than movie clips. OpenAI last year extended a licensing agreement with Shutterstock to use its library of images, videos and music as training data.

But Ridder believes Sora could be used as a tool to enhance the filmmaking process. “It could be a little concerning from an attribution perspective, where someone could use Sora’s work and claim it as their own,” he said. “But maybe filmmakers could use these generated videos to create concept storyboards, rather than using them as the final output for a project.”

Last week, a spokesperson for OpenAI told NBC News in a statement that they would not be "broadly releasing" Sora anytime soon because they wanted to communicate with policymakers, educators, and artists before releasing it publicly.

Like several other generative AI companies, OpenAI has faced multiple lawsuits alleging that it used copyrighted material to train its products. A spokesperson for OpenAI said Sora’s “training data comes from content we have licensed and publicly available content.”

Artists have also expressed concerns about whether Sora will enable studios to generate likenesses of actors without their consent. In a blog post, OpenAI said the text-to-video model will reject text inputs asking it to generate likenesses of celebrities.

Last week, the Federal Trade Commission proposed a rule that would make it illegal to create AI impersonations of real people by expanding existing protections against government and commercial impersonations.

Hollywood has adapted to technological advances before, such as CGI and computer animation software. Mintz hopes the industry can also survive through AI innovation.

“They can’t replicate the creative human mind or make the same creative decisions that artists have been making for hundreds of years. We stand on the shoulders of giants and do our best to create unique work that we can be proud of.”