The core change in this round of Midjourney updates is pushing “generation” toward “controllable editing”: the online image editor is more usable, personalization better matches individual aesthetics, and style reference makes style reuse more consistent. This article, in a hands-on order, clearly explains how to use Midjourney’s new features and what scenarios they fit.
Online Image Editor: Erase/Repaint and Canvas Expansion Feel Smoother
Midjourney’s online image editor focuses on local edits (similar to inpainting) and image expansion, making it suitable for situations where “the overall image is satisfactory but a local area goes off the rails.” A common workflow is to select the problem area first and then revise it—for example, use Erase to remove extra objects and let Midjourney regenerate details in the blank space.
If you erase too much or the selection isn’t precise, you can use Restore to roll that area back and erase again. When making posters or e-commerce images, it’s recommended to expand the canvas first to create the needed negative space, then use local repainting to fix high-frequency problem spots such as text edges, hands, and accessories—the number of rework iterations in Midjourney will drop noticeably.
Personalization: Make Midjourney More Like Your “Default Aesthetic”
Midjourney’s personalization (often referred to as the style tuner/model personalization) is better suited to users with stable preferences: you don’t have to write long style descriptions every time; instead, you “feed” your preferences to the system first. Its value is in reducing style tug-of-war in prompts, making it easier for Midjourney to produce the kind of texture and feel you like.
A practical approach is: start by creating a set of comparison images around the same subject (same composition, different styles), record the directions you prefer, and then gradually solidify them into your own “default taste.” When producing a series of illustrations or a brand visual system, this Midjourney personalization can save more time than repeatedly trialing prompts.


