“Using a Gun” wall-e: Midjourney generates Disney character video in massive copyright lawsuit
Midjourney’s new The AI-generated video tool will produce animated clips that include copyrighted characters from Disney and Universal Studios, including videos of the beloved Pixar character Wall-e with a gun.
It’s been a busy month for Midjourney. This week, Generative AI startup released its sophisticated new video tool V1, which allows users to make short animated clips from generated or uploaded images. The current version of Midjourney’s AI video tool requires images as a starting point. Generating videos using text-only prompts is not supported.
The V1 release comes after another announcement in early June: Hollywood behemoth Disney and Universal Studios filed a sensational lawsuit against Midjourney, claiming it infringes on copyright law by generating images with Studios’ intellectual property rights.
Midjourney did not immediately respond to a request for comment. Disney and the general reiterated statements of executives on litigation work, including Disney’s legal director Horacio Gutierrez claimed that DiJuni’s output amounted to “pirated copies.”
It seems that Midjourney may have tried to put some video-specific guardrails for the V1. In our tests, it blocks prompt-based animations freezingElsa, Boss Baby, Goofy and Mickey Mouse, although it still generates images of these characters. “AI Host” blocked the generated video of the prompt when Wired asked for V1 to animate the image of Elsa. The pop-up message reads: “Al controls the real videos, especially people’s real videos, are cautious.”
These seem to be the limitations of guardrails that are incomplete. Wired tests show that V1 will generate cartoon clips of various Universal and Disney characters, including Homer Simpson, Shrek, Minis, Minis, Deadpool and Star Wars‘C-3PO and Darth Vader. For example, when asked about the image of a minion eating a banana, Midjourney produces four outputs, with a recognizable version of the cute yellow characters. Then, when the wire clicks the “Animation” button on one of the outputs, Midjourney generates a follow-up video where the character is eating bananas (Peel and All).
While Midjourney seems to have blocked some of the video-related Disney and universal related tips, sometimes potential guardrails may be bypassed during testing by using spelling changes or repeated tips. Midjourney also allows users to provide prompts to inform animations; using this feature, Wired is able to produce clips of copyrighted characters acting in an adult way, such as Wall-E wielding a gun and Yoda smoking.
Disney and universal lawsuits pose a major threat to Mijuni, which also faces other legal challenges faced by visual artists who claim copyright infringement. Although it focuses primarily on the examples that provide Midjourney’s image generation tool, the complaint states that the video “only enhances Midjourney’s ability to distribute infringing copies, replicas and derivatives of plaintiff’s copyrighted works.”
The complaints include dozens of alleged Midjourney images showing common characters and Disney characters. The scene was originally part of Midjourney’s report, which is about the so-called “visual theft problem” about AI critic and cognitive scientist Gary Marcus and visual artist Reid Southen.
“Reed and I pointed out this issue 18 months ago, with little progress and little change,” Marcus said. “We still have the same situation with using unlicensed materials and the guardrails are a little effective, but not very good. For all the discussions about index advances in AI, we get better graphics, rather than the fundamentals solution for this problem.”
