Skip to content
Models

Midjourney now restyles images from a moodboard or style code, no prompt needed

Midjourney's edit model now applies a moodboard, style code or style reference to an attached image with no prompt. The inputs were already in its docs, and the demo avoids faces and text.

Drafted by The Generated · Filed by The Generated
Three versions of the same Midjourney image of a winged figure seen from behind on a mountain, each restyled in a different look

Midjourney's edit model now restyles an attached picture from a moodboard, a style code or style reference images when you press enter, with no text prompt needed, the company announced on X on Tuesday 17 November. It showed the change with an 11-second demo that pushes one image through 11 looks.

The useful change is the missing prompt. A saved look now works like a one-click filter on a frame you already have, so a director's board becomes a reusable asset rather than a paragraph of prompt text. The post's wording, repeated by every restatement we read, says Midjourney "added support" for these inputs. Its own help pages have listed them as compatible with the edit model since the summer.

What changed on Tuesday

The post spells out the flow: drop an image into the prompt bar under "attach to prompt", add a favorite moodboard, style code or sref images, and push enter. A moodboard is a saved set of reference images; a style code is a number that recalls a look; sref is Midjourney's style reference parameter. Midjourney did not say which plans get the promptless path, whether it works in Discord, or whether it costs more GPU time than a normal edit.

For filmmakers, the job this does is familiar. Midjourney's Retexture tool used to repaint the surface of an image while keeping its shapes, and the V8.2 release folded it into the edit model on 24 July. Tuesday's change brings that job back as a single step: lock a frame's composition, then push a board's look across stills or keyframes before they go to a video model.

What the demo shows

We watched the 11-second clip frame by frame. One image, a winged figure seen from behind on a mountain, runs through 11 styles: a dusk silhouette, monochrome line art, pastel haze, ink brush, flat color blocks, a primary-color collage, a prismatic glow, a lavender haze, a two-tone print, colored pencil and an illuminated manuscript with script in the margins. The outline, the wings and the stance hold in every frame while the palette and surface marks change.

It is also a best-case input. The figure has its back to the camera, so there is no face to keep, no lettering to preserve and little fine set detail. Asked whether a moodboard can change the surface while keeping the composition, Midjourney replied that the video shows exactly that. That answer is the company's claim from its own demo, and no independent test of faces, text or sets had been published by Wednesday morning.

The weak spot: "added support"

Midjourney's Edit Model help article, created on 28 August and last edited on 24 September, already says the edit model can be used with image prompts, style references, moodboards, personalization and --hd. Its Retexturing section gives the example prompt "hand drawn pencil sketch of this image --sref 2708276657". An independent test write-up from 28 August also listed moodboards and sref as supported at the edit model's launch, with the note that extra prompt direction may help.

So the inputs are not new; dropping the prompt is. The same help article still recommends describing the style in your prompt to help activate the aesthetic, advice Tuesday's post now contradicts, and it had not been updated when we read it on Wednesday. Of the three restatements we read on X, a launch account, a Chinese-language repost and a roundup, all repeat "added support" and none mentions the earlier docs. No trade outlet had written it up when we searched.

Limits and early reaction

Midjourney's docs list a few hard limits that still apply: the edit model does not work with --tile, its results cannot be used with Remix, and the --sw style weight does not work with moodboards, so a board cannot be dialed up or down that way. One user replied that text in an image gets blurred after a restyle, and another said the newer style reference model is worse. Both are single, unverified reports.

The post had 449 likes and 42 replies by Wednesday morning, small against the account's 463,347 followers. Most replies cheer, and several repeat the standing request for an API (a way for other software to call Midjourney directly) or an MCP connector, which Midjourney still does not offer. One illustrator, @vikxel, posted a flat-color drawing restyled with their own moodboard.

What it means for AI creators

  • A look you already built, as a moodboard or a style code, can now be applied to a locked frame in one move, which suits matching stills or keyframes before animation.
  • Test it on what the demo avoided: a face, a sign or title, and a busy set. That is where a restyle tends to drift.
  • Plan access, cost per job and the Discord path are unconfirmed, so check your own plan before building a pipeline on it.
  • There is still no official API, so this stays a manual step inside Midjourney's web app.

Our take

Midjourney has quietly turned a style into something you own and reuse, and that matters more for keeping a film's look consistent than another model version would. The launch line oversells it as new support. The real test is a face in close-up, and Midjourney chose a figure with its back turned.

Timeline

  • Date (2026): Fri 24 Jul. Event: V8.2 becomes the default; the edit model replaces Omni Reference, Character Reference and Retexture; Source: Midjourney Version help article
  • Date (2026): Fri 28 Aug. Event: Edit Model help article created; an independent test lists moodboards and sref as supported; Source: Midjourney Help Center; aireiter.com
  • Date (2026): Thu 24 Sep. Event: Edit Model help article last edited, listing style references and moodboards as compatible; Source: Midjourney Help Center
  • Date (2026): Tue 17 Nov, 17:06 UTC. Event: Midjourney posts that moodboards, style codes and srefs work in the edit model with no prompt; Source: Midjourney on X

Sources and what is unconfirmed: Midjourney's 17 Nov post and 11-second demo video on X (watched frame by frame); Midjourney help articles on the Edit Model, Style Reference, Moodboards and Version, read 18 Nov; aireiter.com test write-up (28 Aug); X search to 18 Nov. Unconfirmed: plan access, cost per job, Discord support and face or text preservation. Midjourney's updates blog is login-walled and was not read. We did not test the feature and did not contact Midjourney.

Your Voice

No prompt needed, just your style. This is how a visual language finally becomes yours.

@RobotCleopatra on X

I’d love to see the side-by-side where the moodboard changes the surface treatment and leaves the composition alone.

@ShivamGuptaim on X

Midjourney’s unmatched aesthetic and style remapped onto any existing image.

@next_on_now on X

Quoted from public posts on X and credited by handle. If one of these is yours and you want it removed, message @the__generated on X and it comes down.

Got something to say? Tag @the__generated on X