Midjourney
Based on Wikipedia: Midjourney
In September 2022, a strange image began circulating online that looked like Pope Francis in a white puffer jacket—but the pope had never worn such a coat. The image was generated by Midjourney, and it went viral in a way that previewed the wild years to come. Within months, the tool would also generate a fictional arrest of Donald Trump, a hoax attack on the Pentagon, and images so real they fooled millions. But none of this was as controversial as what happened at the Colorado State Fair that same year.
A piece called "Théâtre D'opéra Spatial" won first place in the digital art competition—a striking, surreal image of Baroque figures lounging in an ornate theater. The problem: it wasn't made by a human artist. Jason Allen had typed a prompt into Midjourney, printed the result on canvas, and entered it under his name. Other artists were furious. Two category judges hadn't known Midjourney was AI-generated—and when they learned, one said he'd have awarded the prize anyway.
This is the strange world of Midjourney, a generative AI program that creates images from natural language descriptions called prompts. Born from a San Francisco-based company bearing its name as both product and research lab, Midjourney sits in the middle of the artificial intelligence boom reshaping how we think about creativity, authorship, and what it means to make art.
The Birth of a Dream Machine
Midjourney, Inc. was founded by David Holz, who previously co-founded Leap Motion—the hand-tracking technology that once seemed like science fiction. When Midjourney entered open beta on July 12, 2022, the company was already profitable within weeks. By August 2022, Holz told The Register that the young outfit had crossed that rare threshold most startups chase: profitability.
The company launched its Discord server in March 2022, requesting users post high-quality photographs to Twitter and Reddit for training—essentially asking strangers to feed the algorithm. Since then, Midjourney has released new algorithmic versions every few months, each one a step deeper into what text-to-image generation can achieve.
Version 1 dropped in April 2022, followed by Version 3 in July of that year. By November 5, 2022, the alpha iteration of Version 4 emerged—models now trained on Google's Tensor Processing Units, giving them more computational firepower than earlier versions.
The leap to Version 5 came on March 15, 2023, and proved pivotal. The 5.1 model became more opinionated than its predecessor, applying more of its own stylization to images—essentially developing a distinct artistic personality. Users could now choose between RAW mode for literal interpretation or standard modes that added their own flavor.
Version 6 arrived on December 21, 2023, trained from scratch over a nine-month period. It brought better text rendition and more precise prompt adherence—a hallmark of how Midjourney has continuously pushed what generative AI can do.
How to Talk to a Machine That Makes Pictures
Midjourney lives inside Discord, the instant messaging platform popular with gamers and developers. Users generate images using the /imagine command, typing prompts like "cyberpunk city at night, neon rain, cinematic lighting"—and the bot returns four images users can upscale or refine.
Initially, website users needed to have generated at least 1,000 images through the bot before using web tools—a significant gate that has since vanished. Midjourney's Discord server provides public access; users can also invite the bot directly to their own servers for private creation.
In September 2023, midjourney introduced the Vary (Region) feature as part of version 5.2, letting users select specific areas of an image and apply variations only there while keeping the rest unchanged—a powerful tool for precise editing.
Then came August 2024: Midjourney launched its web interface to make tools more accessible beyond Discord. Version 6.1 arrived alongside this shift, consolidating image editing, panning, zooming, region variation, and inpainting into a single interface. The web editor also synced conversations between Discord channels and web rooms—a bridge between the old community and new accessibility.
This move responded to growing competition: Adobe Firefly and Google's Imagen had already launched as native web apps with integration into design tools like Photoshop. Midjourney needed its own space beyond the command line.
Style, Weight, and Character
A pivotal feature lets users control how much influence an uploaded image has on final output—called image weight. Setting a higher weight ensures the generated result closely follows the reference image's structure; lower weights let the text prompt dominate. It's like telling the algorithm: "here's what I want visually" versus "here's what I'm describing."
The Style Reference feature allows uploading an image to use as a stylistic guide—extracting color palette, texture, atmosphere and applying it to newly generated images. This lets creators mimic specific artistic moods without needing technical skill.
Character Reference targets consistency: upload a character image and the system generates similar characters in output—useful for maintaining appearance across different scenes or projects.
Who Uses It?
David Holz told The Register that artists primarily use Midjourney for rapid prototyping—to show clients concepts before the real work begins. It's become valuable in advertising, where tools like this enable custom ads for individuals, new special effects, and more efficient e-commerce campaigns. Ad Age described it as a revolution in original content creation.
Architects have embraced the tool to generate mood boards for early-stage projects—a faster alternative to searching Google Images.
The program reached notable publications: The Economist used Midjourney for its June 2022 front cover, and Italy's leading newspaper Corriere della Sera published a comic created with Midjourney by writer Vanni Santoni in August 2022. Charlie Warzel of The Atlantic generated two images of Alex Jones using Midjourney for his newsletter—though he later called using AI-generated images a mistake.
Last Week Tonight with John Oliver dedicated a 10-minute segment to Midjourney in August 2022, bringing it to mainstream attention.
The Controversies
Midjourney's rise hasn't been smooth. In December 2022, it generated images for an AI-generated children's book titled "Alice and Sparkle," created over a weekend by Ammaar Reeshi—a young girl builds a robot that becomes self-aware. Thirteen images from Midjourney went into the book.
The backlash was severe. One artist noted: "the main problem is that it was trained off of artists' work—our creations, our distinct styles that we did not consent to being used."
By 2023, the realism of AI text-to-image generators reached a threshold that sparked viral waves of misleading photos. A Midjourney-generated picture of Pope Francis wearing a white puffer coat went completely viral—it looked like real documentation but wasn't. The fictional arrest of Donald Trump spread online. A hoax of a Pentagon attack appeared in generated images.
The realism became so convincing that research suggested Midjourney's outputs could be biased—neutral prompts returned unequal results, revealing hidden training biases in the model.
Midjourney stands as both revolutionary and controversial: generating images from words alone, winning art competitions, appearing on magazine covers—and igniting debates about consent, creativity, and what authorship means when anyone can prompt a machine to create visual wonders.