Midjourney V5

What it is

Midjourney V5 is a version of Midjourney's text-to-image model: a system that turns a written prompt into a picture. Users describe what they want — a subject, a style, a mood — and the model generates images to match. V5 was released in alpha on 15 March 2023 ^[1].

How it was released

The rollout was announced a day in advance. On 14 March 2023, Midjourney founder David Holz said V5 would be released that week, and that paying customers would rate alpha images first ^[2]. That ordering shaped the first public impression of the model: before general users saw it, subscribers were already comparing and voting on early outputs. The alpha itself arrived on 15 March 2023 ^[1].

The 5.x line

V5 was not a single fixed release. Over the following months it was revised several times, and each revision adjusted how the model behaved rather than merely how it looked.

V5.1, on 3 May 2023, pushed toward more opinionated stylisation, meaning the model applied a stronger stylistic interpretation of its own. The same release included a RAW variant that follows prompts more literally, giving users a way to choose between a version with a pronounced house style and one that sticks closer to what was typed ^[1].

V5.2, on 22 June 2023, added a new aesthetics system together with a zoom-out feature, which lets a user extend the frame outward from a generated image rather than starting again from a prompt ^[1]. In September 2023 came Vary (Region), an inpainting tool: instead of regenerating a whole picture, a user selects part of it and asks for that area alone to be redone, leaving the rest untouched ^[1].

Alongside the main model, Midjourney ran a separate line with a different visual emphasis. Niji 5 was released on 2 April 2023 ^[1].

Why it drew attention

The period in which V5 appeared was one in which synthetic images became hard to dismiss. In March 2023, a fake image of Pope Francis wearing a white puffer coat, made with Midjourney, spread widely online ^[1]. The picture was not a photograph, but it circulated as though it could have been, and it became a frequently cited illustration of how quickly generated images could travel beyond the context in which they were made.

Where it sits

V5 was a stage in a fast sequence of releases rather than an endpoint. Its successor, V6, arrived in December 2023 ^[1]. Between the alpha in March and that successor, the 5.x revisions added controls — literal versus stylised rendering, zoom-out, region-level editing — that moved the product from producing single images toward letting users direct and revise them.