Toys ‘R’ Us AI-generated ad torn apart by critics

Toys “R” Us has seen a wave of backlash after releasing its latest advertisement, fully created using OpenAI’s generative video artificial intelligence tool Sora.

The ad depicts the American toy retailer’s founder Charles Lazarus as a child and his vision to create the Toys “R” Us store and its mascot Geoffrey the Giraffe. The ad was created by the firm’s in-house studio and creative agency Native Foreign.

In a statement accompanying the ad, the toy retailer touted its AI-generated film as “leaping ahead of the curve” as the first major ad created purely by OpenAI’s text-to-video tool.

“Charles Lazarus was a visionary ahead of his time and we wanted to honor his legacy with a spot using the most cutting-edge technology available,” Toys “R” Us chief marketing officer Kim Miller Olko said in a statement.

The ad was met with criticism from artists and filmmakers, with Avengers: Endgame director Joe Russo saying it “fucking sucks.”

TOYS ‘R US released an AI commercial and it fucking sucks. pic.twitter.com/K1JcGhHKeA

— Joe Russo (@joerussotweets) June 25, 2024

Others lashed the ad from a technical standpoint for its lack of visual continuity, including noticeable changes in the character’s bodily features and clothing throughout the 66-second clip.

“Just like a weird dream it is a different person each time you see them,” one X user wrote.

“This is not the way. Why these marketing people thought this was a good showcase of their product’s soul is baffling,” wrote the CEO of metaverse multimedia firm BasedAF, Robin Schmidt.

Related: OpenAI reportedly considering shift to for-profit as CEO stacks board

Others took aim at the supposed energy costs of AI-generated material compared to conventional filming for promotional material.

“No aspect of this looks better than conventional tools yet it costs an order of magnitude more energy to produce,” former Ubisoft concept artist RJ Palmer said on X.

“Even then, you can still see where figures were crudely cut out from disparate sources to attempt a cohesive whole,” they added. “There is fundamentally no benefit to this.”

OpenAI first unveiled its text-to-video model Sora on Feb. 15, initially wowing users on social media with its capabilities.

However, since its release, limitations of the model including its inability to generate consistent footage without any “uncanny” differences in subjects have come to light in the rare glimpses of generated footage that make it into the mainstream.

X Hall of Flame: Ethereum’s recent pullback could be a gift — Dynamo DeFi