Midjourney will soon launch a web tool for editing images with generative AI, expanding user capabilities while limiting initial access to mitigate risks.
CEO David Holtz announced on Midjourney’s official Discord server that the company plans to release the image editor “early next week.” The tool will enable users to edit images with Midjourney’s AI and retexture them while maintaining shapes and updating colors and details based on prompts. This move expands Midjourney from creating AI-generated art to enabling users to modify existing images.
Midjourney about to drop a new image editor that lets you upload and edit any image.
But, it will be heavily restricted and moderated 😞 pic.twitter.com/xYcTWF36Qe
— Dreaming Tulpa 🥓👑 (@dreamingtulpa) October 19, 2024
Due to the upgraded tool’s powerful features, Midjourney will initially limit access to a “subset of the current community.” Holtz stated that the company will enhance human moderation and introduce “new, more advanced AI moderators” to help prevent abuse.
Holtz acknowledged uncertainty about restricting access to the new feature and mentioned that the company intends to launch a survey to collect user opinions. The findings will be used to assess what percentage of the user base will be affected by different restrictions.
Risks and Regulations of Deepfakes
These advanced AI features reflect trends at Google and Adobe, which have also introduced similar editing tools. However, this raises concerns about distinguishing between AI-modified images and those generated from scratch. Meta, for instance, has struggled to label AI-edited images, underscoring ongoing debates about tracking such edits.
Last year, Midjourney adopted the IPTC’s Digital Source Type property for embedding metadata in AI-generated images but has not embraced C2PA, which tracks an image’s full provenance.
Launching editing tools without proper safeguards poses risks, including large-scale copyright infringement and the proliferation of misleading deepfakes. As deepfakes proliferate on social media, distinguishing between truth and disinformation becomes increasingly tricky, raising concerns for businesses and political entities. This trend is particularly alarming given the US elections, where deepfakes could manipulate public opinion and deceive voters, undermining democracy.
With no federal law against deepfakes, 19 US states have enacted regulations on using generative AI in election communications, while seven more are considering similar bills.
In August, Midjourney launched a website that allows users to generate AI images online without a Discord account. Users can sign up with Google or Discord, explore sections, and create images through text prompts. Subscription plans are available after a free trial.