What will TED look like in 40 years? For #TED2024, we worked with artist @PaulTrillo and @OpenAI to create this exclusive video using Sora, their unreleased text-to-video model. Stay tuned for more groundbreaking AI — coming soon to https://t.co/YLcO5Ju923! pic.twitter.com/lTHhcUm4Fi
— TED Talks (@TEDTalks) April 19, 2024
With a unique first-person view (FPV) drone ride, OpenAI’s upcoming video generation tool, Sora has unveiled a new video that provides an intriguing look into the future of TED Talks. Although Sora’s full release is not expected until later this year, this preview shows how it can revolutionise video creation—particularly by making sophisticated effects more widely available.
Filmmaker Paul Trillo, who has been experimenting with Sora since February, worked with OpenAI, TED Talks, and other partners to create the exciting fly-through of futuristic conference rooms, strange labs, and underwater tunnels. This demonstrates Sora’s artistic potential and its capacity to mimic intricate cinematography techniques, which usually call for costly equipment and highly skilled drone piloting.
But the video also shows off a few of Sora’s present shortcomings. The people’s fleeting appearances and the eerie background elements highlight how difficult it is to represent people realistically and keep a scene cohesive in dynamic scenes. These elements combine to create a dreamlike viewing experience that can be thrilling and a little confusing at the same time.
Paul Trillo gave Business Insider an inside look at his workflow as an alpha tester for Sora. He underlined how crucial it is to use particular cues like “35 millimetres,” “anamorphic lens,” and “depth of field lens vignette” to prevent Sora from producing outputs that appear unduly digital. He compared using Sora to playing slots, where inputs are mixed randomly and mentioned that the lack of a true physics engine in the tool reduces the realism of interactions like breaking glass.
Since Sora does not yet accurately simulate the physics of many fundamental interactions, its use is currently limited to short-form videos rather than full-length films, as noted by OpenAI, which has acknowledged these limitations.
Notwithstanding these difficulties, it is evident that Sora can democratize high-quality video effects and create new avenues for artistic expression. With previews like this creative FPV drone-style video, interest in what OpenAI’s Sora tool is capable of continues to rise as the tool nears its public launch in late 2024.
Rob Williams is a tech enthusiast and digital content creator with a passion for emerging technologies. With over five years of experience in the field of digital marketing and content production, Rob has been closely following the developments in AI and its applications in media creation. When not writing about the latest in AI, Alex enjoys experimenting with new digital tools and sharing insights with the online tech community.