The new generation of the AI of Midjourney The video tool will produce animated clips featuring characters protected by the copyright of Disney and Universal, found Wired, including the video of the beloved Pixar Wall-E Pixar holding a pistol.
It was a busy month to Midjourney. This week, the startup generating AI released Its new sophisticated video tool, V1, which allows users to make short animated clips from images they generate or download. The current version of the AI video tool from Midjourney requires an image as a starting point; The generation of videos using text prompts only is not supported.
The release of V1 occurs in the heels of a very different type of announcement earlier in June: Hollywood Behemoths Disney and Universal have filed a successful trial against Midjourney, alleging that he violates the law on copyright by generating images with the intellectual property of the studios.
Midjourney did not immediately respond to requests for comments. Disney and Universal reiterated the statements made by its leaders on the trial, notably the legal chief of Disney, Horacio Gutierrez, alleging that the production of Midjourney is equivalent to “hacking”.
It seems that Midjourney may have tried to set up specific video railings for V1. In our tests, he blocked the entertainment of prompts based on Frozenis Elsa, Boss Baby, Goofy and Mickey Mouse, although it would always generate images of these characters. When Wired asked V1 to animate the images of Elsa, an “AI moderator” blocked the invitation to generate videos. “Al Moderation is prudent with realistic videos, especially people,” read the context message.
These limitations, which seem to be railings, are incomplete. The wired tests show that V1 will generate animated clips of a wide variety of universal and Disney characters, including Homer Simpson, Shrek, Minions, Deadpool and Star Wars‘C-3PO and Dark Vader. For example, when asked for an image of henchmen eating a banana, Midjourney generated four outings with recognizable versions of cute yellow characters. Then, when cable clicked on the “Animating” button on one of the outings, Midjourney generated a follow -up video with the characters eating a banana – Pel and everything.
Although Midjourney seems to have blocked certain guests linked to Disney and Universal to videos, Wired could sometimes bypass potential railings during tests using spelling variations or repeating the prompt. Midjourney also allows users to provide an prompt to illuminate animation; Using this functionality, Wired was able to generate clips of characters protected by copyright behaving adultly, like Wall-E brandishing a firearm and yoda smoking a joint.
The Disney and Universal trial constitutes a major threat to Midjourney, which also faces additional legal challenges of visual artists who also allege the violation of copyright. Although it is largely focused on the supply of examples of Midjourney’s image generation tools, the complaint lightens that the video “would only improve Midjourney’s ability to distribute copies, reproductions and derivatives of works protected by copyright of complainants”.
The complaint includes dozens of alleged images of Midjourney showing universal and Disney characters. The whole was initially produced in the context of a report on the so-called “problem of visual plagiarism” of Midjourney of the critic and cognitive scientist Gary Marcus and the visual artist Reid Southen.
“Reid and I highlighted this problem 18 months ago, and there has been very little progress and very little changes,” explains Marcus. “We always have the same situation of unlimited materials used, and the railings that work a little but not very well. For all discussions on exponential progress in AI, what we get is better graphics, not a fundamental-prime solution to this problem. ”
