What Is Midjourney? The AI Image Generator That Started on Discord

Midjourney is a generative AI service that turns text prompts into images, and it moved to an open beta anyone could join in July 2022. If you want a one-line answer to "what is Midjourney", it is a bot on the chat service Discord: send it a sentence and four pictures come back. That ease, and a distinctive look, made it one of the first services to put image-generating AI into a great many hands. It was built by a self-funded lab of about ten people in San Francisco.

Sources differ on the open-beta date. English Wikipedia dates the company's announcement that it was "moving to open-beta" to 12 July 2022. A PCWorld article of 26 July reported that the service had opened to everyone for 24 hours on 18 July, and that on 26 July Holz emailed beta testers that the "Midjourney beta is now open to everyone." The first model, V1, dates from February 2022 (per English Wikipedia).

David Holz and a lab of "about 10 people"

Midjourney is led by David Holz, who in 2008 co-founded Leap Motion, a company making a device that tracks hand motion for computer input, and ran it for twelve years. According to an interview The Register published on 1 August 2022, Holz left Leap Motion the year before to start Midjourney. He told the paper the company was "about 10 people", self-funded, with no investors, and already profitable.

The numbers in the same interview give a sense of scale at the time. Hundreds of thousands of people were using the service, which took something like 10,000 servers. About 40 moderators watched the images people made. Most users were having fun, Holz said, but about 30% used it professionally, many of them graphic artists generating variations to show clients before starting work themselves. There were two paid tiers, $10 and $30 a month; PCWorld reported that beta users got about 25 free images to start.

Why Discord

Midjourney began not as an app but as a Discord server. New users joined Midjourney's Discord and typed prompts into newcomers' channels, where the results appeared for everyone in the channel; generated images were public by default. Talking to the bot privately took a paid plan.

Holz explained the choice to The Verge: "people want to make things together." Doing that on iOS would mean building your own social network, he said, whereas Discord already provided one. Someone types "dog", someone else "space dog", someone else "Aztec space dog" — watching other people's prompts and copying them was how people learned to use it.

Image generation arrived in three shapes in 2022. OpenAI's DALL·E 2 was a closed service inviting people from a waitlist in stages; Stable Diffusion released its weights outright; Midjourney was a shared space anyone could walk into. Holz also told The Register that user preferences drove the model's improvement: V3 got much better without adding any more art, simply from data about which images users liked and how they used them.

A state-fair prize and the Pope's puffer jacket

Two images carried Midjourney's name to the general public.

The first came in August 2022, when Jason M. Allen's Théâtre D'opéra Spatial, made with Midjourney, took first place in a digital art class at the Colorado State Fair's art show. That story, and how the US Copyright Office refused registration unless the AI-generated parts were disclaimed, is told in AI art wins the Colorado State Fair.

The second came in March 2023, when an image of Pope Francis seemingly wearing a white puffer jacket went viral. According to BuzzFeed News it was made with Midjourney by a 31-year-old construction worker from the Chicago area. That so many people took it for a real photograph showed how far photorealism had come. Midjourney had released the alpha of V5 on 15 March, shortly before (per English Wikipedia).

Model after model

Midjourney has swapped in new models every few months. The milestones, from English Wikipedia and Midjourney's own announcements, are these.

Announcing the video model, Holz wrote that the eventual destination was real-time, open-world simulation, and that image, video, 3D and real-time models were the building blocks.

On 22 August 2025 Alexandr Wang, Meta's Chief AI Officer, announced that Meta would license Midjourney's "aesthetic technology" for its future models and products. Engadget noted that Meta had already launched its own image generator and video editor, and framed the deal as a way to compete with OpenAI's Sora and Google's Veo.

On the receiving end of lawsuits

Training on images made Midjourney a litigant early. In January 2023 three artists filed a class action against Midjourney together with Stability AI and DeviantArt. What that case alleged and what the court decided in 2024 are covered in the Stable Diffusion article and are not repeated here.

The scale changed in 2025. On 11 June Disney and Universal sued Midjourney for copyright infringement in the US District Court for the Central District of California — NBC News called it the first AI copyright lawsuit from Hollywood giants. The complaint called Midjourney "the quintessential copyright free-rider and a bottomless pit of plagiarism" and alleged that it had made $300 million the previous year. Warner Bros. Discovery sued in September, and the cases were consolidated in November.

According to Variety on 2 July 2026, Midjourney has claimed fair use and argued that the studios engage in the same AI practices, seeking discovery about how they use AI. On 15 June a magistrate judge limited that discovery to consumer-facing AI applications. The court had referred the parties to private mediation with a deadline of 19 August 2026. As of July 2026 there was no judgment. Other fights over AI training and copyright are gathered in the AI and copyright timeline.

Afterwards

Midjourney began in 2022 as a lab of about ten people with no outside capital. Three years later Meta was licensing its technology, and in the same year Hollywood's biggest studios were suing it. Being prized for the quality of its images and being asked what that quality was learned from are happening to the same company at the same time. The wider story is in the AI timeline.