According to the recent report by the American magazine “Variety,” the latest generative artificial intelligence model released by the American AI company OpenAI, named Sora, has recently garnered significant attention both inside and outside the industry. Its powerful ability to generate videos up to 1 minute in length based on text commands has excited professionals in the IT and arts sectors, as well as enthusiasts. However, it has also raised concerns within the entertainment industry, including Hollywood, as the prophecy and related risks of AI and robots replacing human jobs, as depicted in the movie “I, Robot,” have once again become a hot topic.
Higher Video Quality and Realism
”Since OpenAI released the text-to-video generation model Sora last week, people’s predictions about this AI functionality have become increasingly ‘anxious.’ After all, compared to similar tools before, OpenAI seems to have made a huge leap forward,” commented “Variety.” The report states that based on the current video output samples, Sora is the most impressive video generation model to date, with its “realistic effects” expected to play a role in high-end entertainment. However, due to certain limitations, it is currently not capable of handling the scale of Hollywood movie production. Tom Graham, CEO of the AI company Metaphysics, believes, “Today’s creative workers demand comprehensive control over performances and scenes. Models like Sora still have a long way to go to generate Hollywood movies.”
In fact, before the emergence of Sora, there were already AI video generation and editing models with similar functions in the industry, such as Runway and Pika. These models can generate short videos based on text prompts or 2D images, derive animations from static images, repair, replace, or insert new visual elements, and extend shots beyond the original frames, enriching the video content people need.
”Fortune” reported that some Hollywood companies have already applied Runway technology to their projects and received Oscars. In comparison, Sora has achieved some significant upgrades and breakthroughs. For example, the quality and realism of the videos are higher and more lifelike, resembling the real world. The length of the videos has also reached 1 minute, capable of presenting the relevant information needed from the text. It is worth noting that Runway’s second-generation model, launched in August last year, could only generate videos up to 18 seconds each time.
Furthermore, Sora has achieved “temporal consistency,” meaning that even when extending videos, it can provide the model with “foresight,” ensuring that the video theme remains consistent outside the frame. The video example previously released by Sora, titled “Fashionable Women Walking on Busy Streets,” resembles a commercial or movie clip, with characters and scenes showing almost no obvious issues. The movements of the body, details of the hair, and reflections of neon lights appear particularly realistic, which was unimaginable a few years ago.
”AI Storytelling"
"Variety” believes that with its powerful capabilities, Sora has the potential to create movies through AI. These output videos can be stitched together as “shots” to form a movie. However, it is essential to maintain the continuity of characters and scenes. Nevertheless, videos generated based on text prompts and conditional parameters can only create longer “AI stories” if the characters remain consistent in the next scenes.
Currently, the large-scale application of Sora in Hollywood still faces many obstacles. Besides the lack of continuity, this model may “misinterpret” the appearance and behavior in the real world, resulting in visible “physical failures” in images and video output. A recent “failure case” of the Sora model circulating on the internet shows chairs moving as if they were alive, completely beyond the physical possibilities of the real world, and people even “disappearing out of thin air.” Such “mistakes,” besides being laughable, also indicate that there are still many problems to be solved before Sora can create movies.
Since Sora’s video output is based on rendering rather than physical modeling, it cannot provide sufficient controllability for now. Hollywood filmmakers currently cannot grasp and control it. In the short term, introducing this AI model into the creative process will encounter more problems. It requires OpenAI developers to add more software control parameters and improve based on the exposed defects to produce videos comparable to those shot by humans.
Additionally, legal issues such as copyright are also significant reasons why Hollywood cannot directly use Sora. Outputting AI-generated materials to the screen may lead to copyright disputes. Even if AI is only trained using relevant materials, there is still a potential risk of infringement. The key negotiation point of last year’s Hollywood writers and actors’ strike was the scope of application of artificial intelligence, which has not yet been resolved. Questions such as whether the copyright of videos generated by Sora as a creative assistance tool is protected and whether there is any infringement liability for training materials will make Hollywood studios cautious about getting involved in disputes.
Currently, AI tools including Sora may still be used in the early stages of projects, such as quickly generating concept art, character designs, and storyboards to help creators choose directions. However, if an interesting concept character or scene is selected for actual shooting, whether it is for movies, TV shows, or game production, it still faces the risk of copyright claims and copyright protection tracing.
”The Show Begins”
Currently, Sora has not been officially opened to all users. Some industry insiders who have tried it out have marveled at its powerful creativity while also discovering various loopholes and risks of misuse. Just like the false information and rumors caused by Deepfake on the internet before, which caused harm to celebrities like Swift and ordinary people, are questions that OpenAI must face. When Google released the AI video model Lumiere in January this year, it admitted that “there is indeed a risk of creating false or harmful content and being abused when using this technology.”
After Sora is opened to ordinary users and social media creators in the future, a massive amount of video content will inevitably be created. However, not everyone will pay attention to legal issues like creative or advertising professionals. Currently, OpenAI promises to gather feedback from artists, educators, and policymakers globally to understand people’s concerns about this technology.
Although AI products like Sora and Runway still have many areas to improve, the “exponential leap speed” is obvious. Tim Urban, author of “The AI Revolution: The Road to Superintelligence,” predicts based on this, “The 21st century will achieve 1000 times the progress of the 20th century.” The appearance of Sora has also made AI peers and technology companies eager. Kristobal Varenzuela, CEO of Runway, expressed on social media, “The show begins.”