Midjourney New Feature Breakdown: V6.1 Detail Boost, More Accurate Text, and Partial Repainting

Midjourney’s recent key updates focus on the V6.1 model and workflow optimizations: images are more coherent, textures are cleaner, text generation is more reliable, and partial repainting makes “photo-edit-style image output” much smoother. This article breaks down these new features clearly and provides practical, follow-along usage tips.

V6.1 Model: More consistent characters, more pleasing details

What frustrates many people most when using Midjourney is broken limbs and drifting facial features. In V6.1, the structural consistency of arms, legs, hands, and other parts is indeed more stable, with a stronger sense of overall visual unity. At the same time, Midjourney has reduced common pixel artifacts; textures like skin, fabric, and metal are finer, and there are fewer “dirty spots” and less noise in the image.

If you often generate half-body portraits or character designs, in the areas that are “easiest to give it away” such as eyes, small faces, and hands in the distance, V6.1 is more likely than older models to produce a usable draft in a single try.

New Upscaler: Clearer textures, faster output cadence

Midjourney added a new upscaler in V6.1, aiming to raise image and texture quality—especially suitable for “deliverable” posters, concept art, and e-commerce mood images. In practice: after upscaling the same image, detail edges are cleaner and material layering is more apparent.

At the same time, Midjourney’s standard generation tasks are faster as well; the official claim is about a 25% improvement. For people who iterate frequently, this speed-up isn’t just hype—it directly reduces the time cost of waiting and repeatedly queuing.

More accurate text generation: Use quotation marks to lock the words you want displayed

In the past, the biggest problem when making posters with text in Midjourney was “looks like letters but isn’t.” In V6.1, text accuracy is noticeably improved—especially when you put the desired content in quotation marks in the prompt, making Midjourney more likely to generate it as requested.

A recommended approach is to describe the text as a visual element—for example: “Poster title is ‘NEW CITY’, sans-serif font, centered.” Midjourney may still occasionally misspell, but compared with older versions it’s more worth trying for first drafts.

--q 2 mode and partial repainting: Spend quality where it matters

V6.1 adds a new --q 2 mode, which takes longer in exchange for richer texture detail, suitable for materials like leather, fur, fabric, and food. Note, however, that when pursuing texture, Midjourney may occasionally slightly sacrifice local coherence, so it’s more recommended for still life, atmospheric scenes, or close-up shots.

In addition, partial repainting (regional repainting) makes “edit a part without redoing everything” a standard operation in Midjourney: you can select the area that needs changes and repaint only that part, then use prompts to add instructions like “change the fingers to natural five fingers,” “remove the pedestrians in the background,” or “change the collar to a high-neck sweater,” etc. To improve the hit rate, keep the prompt focused on only the repainted part and avoid re-describing the overall style of the whole image, so Midjourney can spend its compute on what you actually want to change.