Midjourney is a generative AI platform known for creating highly detailed and imaginative images from text prompts. Launched in 2022, it has gained popularity among artists, designers, and creators for its ability to produce striking visuals through artificial intelligence.
Midjourney will soon launch an improved web application that lets users utilise its generative AI to modify any submitted photos. The new tool, which CEO David Holtz announced, will also allow users to retexture objects in photographs by changing their colours and details in response to text prompts. It is anticipated to launch early next week.
The emergence of AI-powered picture editing has spurred discussions about how to differentiate between photographs created by AI and those that are edited by it. Such photographs have proven difficult for businesses like Meta to classify, and Google has introduced strong AI editing tools that lack obvious visual cues. Despite Midjourney’s pledge last year to label AI-generated content with metadata embedded in the IPTC’s Digital Source Type property, the company has not embraced C2PA, a technique that traces an image’s entire development history.
A small portion of Midjourney’s community will initially have access to the updated tool, and more human and AI moderation is planned to prevent any potential abuse. Holtz acknowledged that they are unsure of the best way to limit the tool’s distribution and are asking the public for input through a survey.
The introduction of these tools, however, raises worries about copyright infringement and the spread of deepfakes, which are getting harder to stop. According to Clarity, a deepfake detection company, deepfakes are becoming a bigger concern in the wake of recent occurrences, such as the spread of phoney photographs during Hurricane Helene. Their production has increased by 900% since the previous year. According to a YouGov survey, 85% of Americans are concerned about deceptive deepfakes, indicating that public fear is also on the rise.
Federal action is still absent, despite the fact that some U.S. states have passed legislation against AI impersonation. Recently, Midjourney has taken steps to restrict deepfakes, especially in political contexts, after coming under fire for utilizing copyrighted data to train its models.