Toys “R” Us faced a strong backlash following the launch of its latest commercial, entirely crafted using Sora, OpenAI’s generative video artificial intelligence tool. The advertisement portrays the founder of the American toy giant, Charles Lazarus, in his youth, showcasing his dream of establishing the Toys “R” Us store and introducing its beloved mascot, Geoffrey the Giraffe. Crafted by the company’s internal studio and creative agency, Native Foreign, the ad was hailed as a groundbreaking achievement, being the first major commercial produced exclusively through OpenAI’s text-to-video technology.
Kim Miller Olko, the Chief Marketing Officer of Toys “R” Us, emphasized the visionary nature of Charles Lazarus, stating, “Charles Lazarus was a forward-thinker, and we aimed to pay homage to his legacy by utilizing the most advanced technology at our disposal.” Despite this praise, the commercial received sharp criticism from the artistic and filmmaking communities. Joe Russo, the director of “Avengers: Endgame,” bluntly described it as “fucking sucks.” Criticism also arose regarding the visual inconsistencies within the ad, with noticeable alterations in the characters’ physical attributes and attire appearing throughout the 66-second presentation.
Comparisons were drawn to a surreal experience, as one user remarked, “Just like a weird dream it is a different person each time you see them.” The CEO of metaverse multimedia company BasedAF, Robin Schmidt, expressed bewilderment, stating, “This is not the way. Why these marketing individuals believed this was a commendable representation of their product’s essence is perplexing.” Additionally, concerns were raised about the purported energy consumption associated with AI-generated content in contrast to traditional filming methods, with former Ubisoft concept artist RJ Palmer highlighting the inefficiencies and lack of tangible benefits in such endeavors.
OpenAI introduced its text-to-video model, Sora, on February 15, initially impressing users across social media with its capabilities. However, post-release, limitations of the model have come to light, particularly its struggle to create consistent footage without noticeable disparities in subjects—a flaw that occasionally surfaces in the rare instances of generated content that reach broader audiences.