Midjourney External Image Editor & Personalized Models: A Practical Guide to the Latest Update

In this Midjourney update, the two features most worth paying attention to are the “External Image Editor” and “Personalized Models”: one makes it much easier to directly edit and inpaint images, and the other helps Midjourney better match your personal aesthetic. Below, we’ll walk through the key entry points, practical usage, and common pitfalls in a hands-on workflow.

1. What this Midjourney update actually changes

In the past, doing local adjustments in Midjourney often meant generating repeatedly, upscaling, then fine-tuning again—long, and easy to drift off target. Now Midjourney makes “editing” a clearer, more direct workflow: pick an image, select an area, then drive the change with text instructions. The approach feels closer to real photo editing.

More importantly, Midjourney is starting to treat “the style you like” as a trainable preference, rather than something you have to prompt from scratch every time. This means the more thoughtfully you vote in Midjourney, the fewer prompt details you may need later.

2. External Image Editor: bring your own images into Midjourney for edits

The core value of the External Image Editor is that you’re no longer limited to editing “images generated by Midjourney.” You can also upload your own photos, product images, or sketches and have Midjourney extend or redraw them. Common use cases include: filling in backgrounds, changing local materials/textures, or swapping elements while keeping the overall composition.

For a smoother workflow, use the web version: log in to the Midjourney website and open the editor entry (usually visible on the image details page or an edit page), then upload your image → use a selection/mask to mark the area you want to change → in your prompt, clearly state what must stay the same and what should be replaced. For more consistent results, write constraints in detail—for example: “keep the subject’s pose and the direction of light, only replace the clothing material with linen, off-white color.”

3. How to enable Personalized Models: vote in Rank first, then add --p to your prompt

Midjourney’s Personalized Model isn’t something you can simply toggle on in settings. You first need to vote on preferences on the official Rank page. The official guidance notes that after you complete at least a certain number of image votes, you can add --p at the end of your prompt to apply personalization.

In practice, it’s best to do a focused round of voting first to stabilize what you truly like (photographic look, illustration feel, clean negative space, strong contrast, etc.). Then, for the same prompt, compare results “without --p” vs. “with --p.” If the style becomes more consistent, personalization is likely taking effect.

4. Web browsing & output management: use scroll-to-browse to speed up selection

Many people miss a major “viewing efficiency” boost on Midjourney web: when you open a single image in the gallery, you can quickly switch through images with the scroll wheel, significantly reducing paging—especially useful when choosing a final version from many similar compositions. You can quickly shortlist three candidates first, then return to the editor for localized redraws, avoiding wasted attempts on weaker options.

If your workflow is “generate → pick → edit → generate again,” it’s worth treating the Midjourney web app as your main workspace: review results in one place, do edits in one place. Discord becomes more of an optional entry point rather than a required step.

5. Common pitfalls: prompt structure and result consistency

When editing with Midjourney, the most common issue is “I changed A, but B changed too.” The solution isn’t to blindly add more words—it’s to split your prompt into two parts: first list what must remain fixed (subject, pose, lighting, camera), then list what should change (region, material, color, element replacement).

Also, Personalized Models can make Midjourney align more closely with your preferences, but it may also push some commercial work toward an overly strong style. If that happens, run a neutral version first by removing --p, then use that as a base in the editor for subtle adjustments—it tends to be more stable.