Midjourney

In 2021, Leap Motion co-founder David Holz left that company and founded Midjourney in San Francisco. Holz said he wanted to leave the environment of a big venture-backed company and find a “home” for the next ten years to work on cool projects that matter; the company has no investors and runs on its own funds.⁠[1][2][2][3][3]

On July 18, 2022, Midjourney opened to everyone for 24 hours; afterward Holz announced in an email to beta testers that the beta was open to everyone. Users had to go through the chat app Discord: they typed text prompts in Midjourney’s Discord server and received generated images in about a minute, visible to everyone else in the same channel. Beta users could make about 25 free images and then subscribe at $10 a month (200 images) or $30 a month (unlimited), with paying users able to message the bot privately to generate images.⁠[4][4][1][4][1]

In August 2022, Midjourney had about 10 staff and no investors; Holz described it as a self-funded research lab that was already profitable, used by hundreds of thousands of people at the time. Generated images were public by default, and the terms of service banned adult content and gore and asked users to avoid visually shocking or disturbing content.⁠[3][1][1][1][1]

In August 2022, game designer Jason M. Allen won first place and a $300 prize in the emerging artist division’s “digital arts/digitally-manipulated photography” category of the Colorado State Fair Fine Arts Competition with “Théâtre D’opéra Spatial,” an image made with Midjourney. After he shared the news on Midjourney’s Discord server on August 25, it went viral on Twitter and angered many artists; artist Genel Jumalon called it “pretty fucking shitty,” and one user wrote, “We’re watching the death of artistry unfold before our eyes.”⁠[5][5][5][5][5][6][6]

Allen responded that it took him more than 80 hours, 900 iterations, and Photoshop cleanup to produce his three entries, and urged people to recognize the technology as a powerful tool and use it for good rather than hating it. Cal Duran, an art teacher who was one of the judges, said he had not realized the work was AI-generated when judging but stood by the first-place award.⁠[5][5][5][5]

In January 2023, artists Sarah Andersen, Kelly McKernan, and Karla Ortiz sued Stability AI, Midjourney, and DeviantArt, alleging that the companies had infringed the rights of “millions of artists” by training AI tools on five billion images scraped from the web without the original artists’ consent. The suit was brought by lawyer Matthew Butterick with the Joseph Saveri Law Firm, and the complaint was also criticized for technical inaccuracies. Developers of generative AI tools generally argue that training on copyrighted data is covered by fair use in the United States.⁠[7][7][7][7]

On March 15, 2023, Midjourney released version 5 of its model. Earlier Midjourney images were relatively easy to spot because of flaws such as hands with too many fingers; V5 brought marked improvements in details such as hands, lighting, and fabric, making images more realistic.⁠[8][8][8]

In March 2023, fake images made with Midjourney, including Donald Trump being arrested, Pope Francis in a white puffer coat, and Emmanuel Macron walking through protests in Paris, spread widely online and were taken as real by many people. On March 28, Midjourney halted its free trial, which had allowed 25 free images a month. Holz said the halt was because users signing up for multiple free trial accounts were bringing down the service for paying users and that the company was still figuring out the right moderation policies; it later tried reopening trials with new abuse safeguards but closed them again when those proved insufficient.⁠[8][8][8][8][8][8][8]

Midjourney allowed images of world leaders such as Joe Biden and Vladimir Putin but not of Xi Jinping, along with other unpublished blocked terms such as “Afghanistan,” and after the fake Trump arrest images some depictions of arrests were also restricted. Holz explained on Discord that the company just wanted to minimize drama, that “political satire in china is pretty not-okay,” and that “the ability for people in China to use this tech is more important than your ability to generate satire.”⁠[9][9][9]

On March 31, 2023, Sarah McLaughlin, a scholar at the Foundation for Individual Rights and Expression, criticized the rule, arguing that it applied to users everywhere, including the United States, effectively exporting an authoritarian state’s censorship rules worldwide through a tech company.⁠[9][9][9]

On July 30, 2024, Midjourney released its V6.1 image model, improving the coherence and quality of bodies, hands, plants, animals, and other subjects, making standard jobs about 25% faster, improving the accuracy of text placed in quotation marks in prompts, and introducing a new personalization model.⁠[10][10][10]

On August 12, 2024, Judge William H. Orrick of the U.S. District Court for the Northern District of California reviewed the artists’ amended complaint, rejected Midjourney’s argument that the plaintiffs had failed to identify each allegedly infringed work, and found the copyright infringement claims against Midjourney sufficient to proceed; the plaintiffs also alleged that Midjourney used artists’ names with its software, misleading consumers into believing they endorsed it, and trained it to recreate the distinctive look of their work, and the judge found these trademark and trade dress claims sufficient to proceed as well.⁠[11][11][11][11][11]

On August 23, 2024, Midjourney opened its web platform to all users, who could sign in with a Discord account to keep the history of images made on Discord or merge Discord and Google accounts.⁠[12][12]

On April 4, 2025, Midjourney opened an alpha test of V7, which understood text and image prompts better, produced higher image quality, rendered bodies, hands, and other details more coherently, and was the first model with personalization turned on by default. It launched alongside Draft mode, which cost half as much and rendered ten times faster, allowing prompts to be revised conversationally or by voice on the web.⁠[13][13][13][13][13]

On June 11, 2025, six companies including Disney, Marvel, Lucasfilm, 20th Century, and Universal sued Midjourney for copyright infringement in federal court in Los Angeles, calling it a “bottomless pit of plagiarism.” The complaint said that with a simple prompt Midjourney would generate high-quality, downloadable images of copyrighted characters such as Darth Vader, Yoda, the Minions, and Shrek, and that Midjourney had used tools such as crawlers to download content from the internet to train its models. Disney chief legal officer Horacio Gutierrez said AI can be used responsibly, “but piracy is piracy.”⁠[14][14][14][14][15][15]

On June 18, 2025, Midjourney released the first version of its video model to the whole community, using an “Image-to-Video” workflow in which users make images as usual and press “Animate” to set them in motion. The company said the ultimate goal of the technology is real-time open-world simulation that can be moved through and interacted with in 3D, requiring image, video, 3D, and real-time models that would first be released separately and then gradually combined.⁠[16][16][16][16]

On August 22, 2025, Meta chief AI officer Alexandr Wang announced a partnership with Midjourney to license its “aesthetic technology” for Meta’s future models and products. Wang said that delivering the best products required working with the best players in the industry and praised Midjourney’s technical and aesthetic accomplishments.⁠[17][17][17]

In its response, Midjourney argued that training generative AI on copyrighted images is fair use and that the platform is “an instrument for user expression” that assists in creating images only at users’ direction; it also noted that the two studios themselves use generative AI tools including Midjourney, arguing that they “cannot have it both ways,” profiting from industry-standard AI training practices while accusing Midjourney of wrongdoing for the same.⁠[15][15][15]

On September 4, 2025, Warner Bros. Discovery sued Midjourney in federal court in Los Angeles, alleging that it trained its models on the company’s works without consent and let users generate images and videos of characters such as Superman, Batman, Wonder Woman, Tweety, and Scooby-Doo, saying Midjourney “thinks it is above the law” and had made a profit-driven decision to offer copyright owners zero protection despite knowing the scope of the infringement. The complaint noted that Midjourney already had a moderation system blocking violent and sexual content and could likewise screen copyrighted characters but chose not to; it sought statutory damages of up to $150,000 per work and an injunction.⁠[18][18][18][18][2][2]

The complaint said Midjourney’s 2024 revenue reached $300 million and that it had nearly 21 million users as of September 2024. Midjourney did not immediately respond to a request for comment.⁠[2][18]

In November 2025, Midjourney signed a licensing agreement with handheld ultrasound device maker Butterfly Network, securing exclusive rights to its ultrasound-on-chip technology.⁠[19]

On March 17, 2026, Midjourney opened an early version of V8 for testing on its alpha website. V8 followed detailed directions better, rendered text in quotation marks better, generated images about five times faster, and added an --hd mode that natively renders at 2K resolution; on July 24, 2026, it released V8.2, focused on aesthetics, image quality, and personalization.⁠[20][20][20][20][21]

On June 15, 2026, in the lawsuits brought by Disney, Universal, and Warner Bros. against Midjourney, Magistrate Judge Joel Richlin ruled that the studios needed to provide information only about consumer-facing AI applications; in July, Midjourney asked Judge John Kronstadt to overturn that decision and sought the three studios’ internal AI business plans, training datasets, model weights, board materials, and more. Midjourney’s attorney Bobby Ghajar argued that if the plaintiffs were doing the very thing they sought to punish, the evidence went to the heart of its fair use and unclean hands defenses. The studios’ lawyer David Singer called it a “fishing expedition” and said the studios did not seek to stop AI technology or shut down Midjourney’s business but wanted it to stop copying their films and characters.⁠[22][22][22][22][23][22]

On June 17, 2026, at a livestreamed event in San Francisco, Midjourney unveiled its first hardware project, the full-body ultrasound scanner Midjourney Scanner, and created a Midjourney Medical division. A person stands on a platform that lowers them into water at about five centimeters per second, passing through a ring of about half a million tiny elements that emit and record ultrasonic waves, producing a sub-millimeter 3D image of the body in about 60 seconds; each machine uses 40 of Butterfly Network’s ultrasound-on-chip modules, the imaging itself relies on ultrasound and signal processing, and AI is used only to segment and label scan results.⁠[24][25][25][19][19][25][25][25]

Over the following 12 months the company planned research trials and second-generation hardware design, sought FDA approval for diagnostic capabilities, planned to open a “Midjourney Spa” with scanners near Union Square in San Francisco around the end of 2027, aimed to launch a third-generation machine with custom silicon in 2028, and hoped to deploy more than 50,000 scanners worldwide by 2031. The project is led by Head of Medical Tom Calloway.⁠[19][19][25][25][19]

Ana Maria Constantin, writing for The Next Web, questioned the scanner’s claims: Midjourney had never built a physical product or operated a medical device, the scanner had no regulatory clearance, and it currently produced only body composition maps; she also noted that whole-body screening of healthy people is contested in medicine because false positives and incidental findings lead to unnecessary follow-up procedures. Midjourney said it had not yet obtained the FDA approval needed for diagnostic use and intended to pursue it over time.⁠[25][25][25][25]

On July 23, 2026, Midjourney announced its first acquisition, Co–Star. Co–Star remains entirely under the control of its founder, Banu, who with her team will help Midjourney build its product and design organization; Midjourney also said it would announce about half a dozen similarly ambitious projects over the next six months.⁠[26][26][26][26]

On July 30, 2026, automotive photography company EVOX Productions sued Midjourney in the U.S. District Court for the Central District of California, alleging that it downloaded at least tens of thousands of copyrighted car photos from online datasets without permission and stripped watermarks and copyright information before using them for training; EVOX had filed similar suits against Stability AI and Google. Midjourney did not immediately respond.⁠[27][27][27][27]

On October 8, 2026, Midjourney began testing a “Thinking Mode” for image generation on its alpha website, saying it helps improve prompt accuracy, typography, and coherence.⁠[28][28]

As of October 2026, Midjourney describes itself as a community-funded research lab of about 60 people, a lean, self-funded, distributed team with no outside investors; it describes its mission as exploring new mediums of thought and amplifying the imaginative powers of the human species, and says it will release image and video models, software, medical, and hardware projects under themes of imagination, coordination, reflection, beauty, and human flourishing.⁠[29][29][25][30][29]

Midjourney: Official website: https://www.midjourney.com/ (opens in a new window).⁠[30]

X: Public account: https://x.com/midjourney (opens in a new window)⁠[30]