Skip to content
← All articles

What is Midjourney

What it is Midjourney is a generative AI service that creates images from natural-language prompts. It is known for a distinctive cinematic, painterly aesthetic that made it the favorite of concept a

Illustration generated for Neuricon.

Oct 13, 2026 · 4 min read

#models#explainers

What it is

Midjourney is a generative AI service that creates images from natural-language prompts. It is known for a distinctive cinematic, painterly aesthetic that made it the favorite of concept artists, designers, and illustrators, and the default tool for anyone who wanted beautiful images without learning prompt engineering. The company, Midjourney Inc., is a San Francisco-based self-described independent research lab founded and led by David Holz, previously a co-founder of Leap Motion.

The open beta launched on July 12, 2022, after the Discord server opened that March. The company's structure is unusual in the AI industry: it took no venture capital, became profitable within a month of the beta, and by 2025 was discussed at an implied valuation around $10 billion on subscription revenue alone, with a team of only one to two hundred people. It is the rare AI lab that grew the old-fashioned way, on revenue.

How it works

Users prompt a Discord bot, or since August 2024 a web interface, with an imagine command. The service returns four candidate images, which the user can upscale, vary, or edit. Behind the scenes it is a proprietary diffusion-based text-to-image model; the architecture was never published. You rent access to the compute; unlike Stable Diffusion, there are no weights to download.

The Discord-first design shaped the culture. Generations happened in public channels where everyone could see everyone else's prompts and results, which turned prompting into a spectator sport and accelerated the community's learning. Parameters control aspect ratio, stylization, chaos, and version, and the community developed a shared vocabulary of modifiers for lighting, lens, and film stock.

The model lineage moved fast: V1 in February 2022, V2 that April, V3 in July, V4 that November, V5 in March 2023 with a leap in photorealism, V5.1 and V5.2 through mid-2023, V6 in December 2023, V6.1 in July 2024, V7 in April 2025, and V8 in March 2026. Versions from V4 onward trained on Google TPUs. Newer capabilities include inpainting, outpainting, image prompts, style and character references, a blend command for merging images, and still-to-video generation.

Key details

Midjourney is subscription-only, and the free trial was removed in July 2023 after abuse. Reported tiers were Basic at $10 per month for around 200 generations, Standard at $30 with fast and relaxed generation hours, Pro at $60 adding stealth mode for private generation, and Mega at $120. These figures come from third-party trackers, so check the company's current pricing before relying on them. The usage model is generation-based rather than token-based, which makes costs predictable for heavy users.

Limitations

Midjourney has faced sustained criticism on three fronts. Studies found biased outputs on gender, skin color, and location even from neutral prompts, and researchers found it easy to generate racist and conspiratorial imagery. The default aesthetic leans Western and the model has a reported tendency toward national stereotypes.

The copyright lawsuits piled up. An artists' class action arrived in January 2023. Disney and Universal sued in June 2025 over what they called a bottomless pit of plagiarism. Warner Bros. Discovery sued that September over its characters, and further suits were consolidated through 2025 and 2026. The company has said little publicly about its training data, which is the core of every complaint.

Content moderation has been controversial too. Before May 2023, a banned-word system blocked terms including the names of religious figures and political leaders, which users found arbitrary; the company then moved to AI-powered moderation. And the technology's misuse for misinformation is well documented, from the viral fake of the Pope in a puffer jacket in 2023 to a Midjourney-generated anatomical diagram that contributed to a journal paper's retraction in 2024. Beautiful output and unresolved questions about how it was made remain the two defining facts about the service.

Versions and the road to V8

Each Midjourney version had a distinct personality, and the community learned to talk about them the way photographers talk about film stocks. V4 in late 2022 was the first version that felt like Midjourney rather than a generic image model, with its painterly coherence. V5 in March 2023 made the leap to photorealism so convincing that its outputs started appearing in misinformation panics. V5.1 and V5.2 refined the aesthetic through mid-2023, and V6 in December 2023 finally cracked legible text rendering, the weakness that had plagued every image model.

V6.1 arrived in July 2024 alongside the web interface, which freed the service from Discord and opened it to users who had never touched the chat app. V7 followed in April 2025 as an alpha, and V8 in March 2026, each pushing fidelity and prompt understanding further. The versioning cadence, roughly two major releases a year, became the rhythm the community planned around, with alpha releases generating the kind of excitement hardware launches once did.

The prompt culture deserves mention. Because generations were public on Discord, users learned from each other's prompts in real time. A shared vocabulary emerged for controlling the output: aspect ratios, stylization values, chaos parameters, seeds for reproducibility, style references for consistency. Prompt engineering for images was arguably invented in those public channels, and the techniques migrated to every other image tool. Midjourney never published its model, but it published a generation of image-makers.

Comments

Loading comments…

Leave a comment

Posting as (guest — sign in for a profile)