Midjourney's new AI has been generated The video tool generates animated clips featuring Disney and Universal copyrighted characters.
It's been a busy month for Midjourney. This week, Generative AI Startup released its new, sophisticated video tool V1. This allows users to create short animation 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 release of V1 comes just after a very different kind of announcement at the beginning of June. Hollywood giants Disney and Universal have filed a blockbuster lawsuit against Midi Joanie, claiming they violated copyright laws by generating images with studio intellectual property.
Midjourney did not immediately respond to requests for comment. Disney and Universal have repeatedly repeated statements made by executives about the lawsuit. Disney's legal head Holocio Gutierrez claims that the output of the Mid Journey is equivalent to “pirates.”
It appears that Midjourney may have tried to set up a video-specific guardrail for the V1. In our test, we blocked animation from the prompt based on It's frozenElsa, Boss Baby, Goofy and Mickey Mouse generate images of these characters. When Wired asked V1 to animate an image for the ELSA, the “AI Moderator” blocked the generation of the video. Read the pop-up message: “Al Moderation is especially cautious about real videos of people.”
These restrictions, which look like guardrails, are incomplete. Wired tests show that V1 generates animated clips of various Universal and Disney characters, including Homer Simpson, Shrek, Minion, Deadpool, and Deadpool. Star Wars'C-3PO and Darth Vader. For example, when asked for an image of a banana-eating minions, Midjourney produced four outputs with a recognizable version of the cute yellow character. Then, when Wired clicked the “Animation” button in one of the outputs, Midjourney produced a follow-up video, eating peels and more, as the character was eating a banana.
It appears that Midjourney has blocked some of the video's Disney-universal related prompts, but Wired can avoid potential guardrails during testing by using variations of spelling and repeating the prompts. Midjourney also allows users to provide prompts to notify them of animations. Using that feature, Wired was able to generate clips of copyrighted characters that act in an adult way, such as Wall-E swinging a firearm around and Yoda sucking on the joints.
The Disney and Universal lawsuits pose a major threat to the mid-journey, which also faces additional legal challenges from visual artists claiming copyright infringement. Although it focuses primarily on providing examples from Midjourney's Image Generation Tool, the complaint argues that it “enhances Middi Joanie's ability to distribute infringing copies, reproductions, and derivative works of plaintiffs' copyrighted works.”
The complaint includes dozens of midi Joanie images showing Universal and Disney characters. The set was originally created as part of a report on Midjourney's so-called “visual plagiarism issue” from AI critic and cognitive scientist Gary Marcus and visual artist Reid Southen.
“Reed and I pointed out this issue 18 months ago, but there was little progress and little change,” says Marcus. “We have the same situation where unlicensed materials are used, and it works a little bit, but it doesn't work that well. All the talk of exponential advances in AI is not a basic principled solution to this problem, but a better graphics.”
