Midjourney V6

Overview

Midjourney V6 is a version of Midjourney's image generation model, released in alpha on 21 December 2023 ^[1]. "Alpha" is a common software label for a version that is public but still being revised while people use it. According to Midjourney, V6 was trained from scratch over nine months ^[1] — rather than being a further adjustment of the previous version's parameters, it was a newly trained model.

The version number matters because Midjourney ships its image model in numbered generations, and each generation changes what the model is good at, how fast it runs, and which editing tools work with it.

Text inside images

One of the most noticed changes in V6 was improved rendering of text in images ^[1]. Lettering is a long-standing difficulty for image generators: a model builds a picture by producing pixels, and letters demand precise strokes, even spacing and correct spelling, which such models easily blur or garble. Better text rendering makes images that carry signs, labels or titles more usable.

V6.1

V6.1 was released on 30 July 2024 ^[2]. Its focus was coherence: limbs, hands, bodies, plants and animals were rendered more coherently ^[2]. Hands and limbs are among the places where image models most often go wrong. The update also added new 2x upscalers, tools that enlarge a finished image while adding detail ^[2].

On speed and control, V6.1 made standard image jobs about 25% faster, rendered words in quotation marks more accurately, and made personalization codes versioned ^[2]. A personalization code is a marker tied to a set of images a user has picked to steer the model's style; versioning means codes from different model versions no longer overwrite each other.

V6.1 also introduced a new personalization model and a --q 2 mode: jobs in that mode took 25% longer and could sometimes add more texture, at the cost of reduced coherence ^[4]. Not every capability was upgraded with it — V6.1 did not include a new inpainting or outpainting model, so zoom, reframe, repaint and vary region fell back to the V6.0 model ^[4]. Inpainting means redrawing only a selected area of an image; outpainting means extending the picture beyond its original edges.

V6.1 became the default model for all users on release, and Midjourney said it would likely ship a V6.2 within about a month ^[4].

Website interface

In August 2024, alongside V6.1, Midjourney launched a website interface for its image generator ^[1][3].

Alongside and after

Niji 6 was released on 29 January 2024, during the V6 period ^[1]. The successor to the V6 line, V7, arrived in April 2025 ^[1]. Taken together, the sequence — an alpha in December 2023, Niji 6 the following January, a 6.1 update with a website in mid-2024, and V7 in 2025 — shows a model line revised in small, frequent steps rather than in a single release.

Why it mattered

V6 is best understood as the generation in which Midjourney's model became noticeably better at two things users ask for constantly: legible text inside pictures, and believable human and animal anatomy ^[1][2]. The 6.1 update added a second theme: the surrounding tooling. Faster standard jobs, versioned personalization codes, larger upscalers and a browser-based interface ^[2][3] are not changes to the picture-making itself, but they change how the model is used day to day. The fact that zoom, reframe, repaint and vary region stayed on the V6.0 model in V6.1 ^[4] also shows how such releases work in practice: the main generator and the editing tools are updated on separate schedules.