Midjourney is a text-to-image generative AI built by an independent, self-funded research lab founded by David Holz. According to Midjourney’s own documentation, the service turns a written prompt into a set of images: a user types a description and the system generates a batch of pictures to choose from and refine. For its first years Midjourney was used primarily through a Discord server, with image creation later moving to the midjourney.com website.
Midjourney opened a public beta in July 2022 and grew quickly into one of the most widely used image generators, alongside DALL-E 2 and Stable Diffusion, all of which arrived in the same period. Its outputs became known for a distinctive, painterly aesthetic that many users found more immediately striking than competing tools. The product has advanced through successive model versions, each generally improving photorealism, coherence, and prompt adherence; Midjourney’s documentation is the live reference for the current version and feature set, which this entry does not freeze.
Distribution is through Midjourney’s own subscription service rather than open weights or a general developer API.
Note on sourcing: Midjourney is an unusually private company that publishes little beyond its product and documentation sites, and docs.midjourney.com returns an HTTP 403 error to automated fetchers. The descriptive facts above are drawn from Midjourney’s own documentation as captured in the existing 2022 Midjourney milestone entry (verified the same day) rather than from a fresh fetch. Specifics that Midjourney does not publish durably - exact model-version release dates, parameter counts, and training details - are not first-party-sourceable here and are therefore not asserted.
Why business readers should care: Midjourney shows that a small, self-funded team could build one of the most popular and commercially successful image generators without venture backing or open weights, competing on output quality and a recognizable style rather than scale or openness.