On Monday, a video was shared by Will Smith on his official Instagram account, parodying an AI-generated clip of him eating spaghetti that gained popularity last year. The emergence of OpenAI’s Sora video synthesis model has sparked discussions about the significant advancements in AI-generated video quality within the past year, particularly in comparison to the infamous spaghetti video. In his latest video, Smith humorously exaggerates the idea by depicting himself actually eating spaghetti in a comedic manner while jokingly attributing it to AI technology.
Titled “This is getting out of hand!”, the Instagram post features a split-screen layout. The top portion showcases the original AI-generated spaghetti video produced by a Reddit user named “chaindrop” in March 2023, labeled as “AI Video 1 year ago.” Below, in a section titled “AI Video Now,” Smith presents 11 video segments of himself engaging in exaggerated spaghetti-eating actions such as slurping, using his fingers as utensils, and playfully nibbling on a friend’s hair. The background music for the video is “Snap Yo Fingers” by Lil Jon.
In the comments section of the Instagram post, some users expressed confusion regarding the authenticity of the new video, with one individual stating, “I’m still in doubt if the second video was also made by AI or not.” Another commenter humorously remarked, “Boomers are gonna loose [sic] this one. Second one is clearly him making a joke but I wouldn’t doubt it in a couple months time it will get like that.”
While there hasn’t been an attempt yet to create a new AI-generated video of Will Smith eating spaghetti using the Sora model, it is anticipated that such an endeavor would yield significantly improved results compared to the previous year, even if minor glitches were present. With the rapid advancements in this technology, it is plausible to expect that by 2025, video synthesis AI models could replicate the parody video crafted by Smith himself.
It is essential to acknowledge, for historical reference, that the spaghetti-eating video of Will Smith did not exemplify the cutting edge of text-to-video synthesis technology at the time of its creation in March 2023 (a distinction that likely belonged to Runway’s Gen-2, which was undergoing closed testing). Nevertheless, the spaghetti video was relatively sophisticated for open weights models during that period, leveraging the ModelScope AI model. While more advanced video synthesis models were available at the time, the humorous and culturally relevant nature of the spaghetti video makes it a compelling point of comparison for assessing the progress in today’s AI video synthesis capabilities.
