Muse Image

byMeta Superintelligence Labs

Posters, charts and QR codes with type that spells right, plus edits that change only the one thing you asked for

Render the brief
Night harbour grain silo with a white projected tide chart showing two high tides across 00–24 hours, labelled TIDE · METRES and HIGH 4.2 m, reflected in black water

How Muse Image works

Brief it like a layout, not a mood: quote the words that must appear, name what each connected image contributes, and chain follow-up edits on the result.

Muse Image flow in Fuser: a marble bust photo and a lifeguard-chair photo wired into the node, output showing the bust in a straw hat on the chair

Write the layout brief

Describe the scene and layout, put headlines in quotes, or phrase one clear edit and name what must stay unchanged.

Muse Image node settings: edit prompt naming @Image1 and @Image2, two connected images, 3:4 aspect ratio, PNG output

Attach reference images

Connect up to ten images as subject, setting or style sources, and refer to each by input order in your prompt.

Muse Image output: a weathered marble bust in a straw hat on a wooden lifeguard chair under a turquoise sky, with a sign reading ON DUTY 10-18

Set ratio and render

Pick a ratio from 21:9 to 9:21 or leave Auto, choose PNG, WebP or JPEG, then render.

What Muse Image is good at

Built for images that have to say something exactly: legible type, believable charts, scannable codes, and edits that leave everything else alone.

Greengrocer's chalkboard bar chart of lemons sold Monday to Saturday, each bar made of real lemons taped up, with a QR code card in the corner

Labels that spell right

Quote short strings and headlines, axis labels and price cards come out correctly spelled most of the time. Bars are drawn by eye, not to scale, so for exact data or a QR code that scans, connect the real chart or code and let the model restyle it.

Impressionist oil painting of a willow-lined river with a rowboat in the reeds and a row of upright pianos floating in the sky, reflected in the water

Multi-reference scene composition

Switch to edit mode and combine up to 10 images, taking the subject from one and the setting from another. Name each by input order and ask the model to match the street's lighting.

One-change precision edits

Edit mode regenerates the full image yet holds lighting and background steady, so a garment color swap or an added sign changes only that element. Chain edits by feeding each output back in.

Two ferry screenprints laid on a pale ash plan chest: one wide panoramic 21:9 print and one tall narrow 9:21 print

Ratios from 21:9 to 9:21

Ten aspect options include Auto, which lets the planner choose framing for posters and infographics. Lock a ratio when the image must slot into a fixed layout or match a source frame.

Made with Muse Image

A gallery of literal, layout-faithful frames: a magazine cover, a product hero, an architectural concept, a film still and a record sleeve, each carrying small text that holds up.

Fashion campaign: woman in a chartreuse coat striding past an ultramarine wall on a red floor, small Look 14 caption

Editorial fashion campaign with electric color contrasts

Invented late-1960s desk fan in warm grey plastic with a clear acrylic 0–3 speed wheel and an ARIA 3 nameplate on a chrome tubular stand

Industrial design hero capture

Architectural concept of a travertine market hall whose concave facade folds into broad public stone steps, inscribed MARKET HALL · 2027

Architectural concept visual

Coyote on desert rocks at dusk holding an iridescent foil balloon in its mouth

Cinematic film frame

Album cover: the word HALOREN blind-embossed into moss-green sand texture, smearing like wind-blown sand, with LP · SIDE A debossed below

Album art graphic layout

What people build with Muse Image

Reached for by designers and marketers whose images carry copy, numbers or a product, and by anyone art-directing one controlled change at a time.

Poster and gig-flyer designers

01

Get a headline and date that actually read, laid out where you said. Short quoted strings and explicit placement make poster drafts usable on the first or second try.

Packaging and ad teams

02

Mock up labels, signage and packaging with crisp small type, then rerun with a new line of copy without redoing the scene. Connect a real QR code when it has to scan.

Infographic and data editors

03

Build the chart in code for exact figures, then connect it and have Muse restyle it into a clean editorial graphic that keeps the bars, numbers, axis labels and titles.

E-commerce colorway editors

04

Swap a garment color, change the backdrop or add a tag across a catalog shot while the model, pose and lighting stay locked, one instruction at a time.

Scene builders from references

05

Place a character from one image into the setting of another and keep the lighting believable, ideal for storyboard frames and concept stills.

Muse Image specs

Plans and credits
Inputs
Prompt (text), Images (up to 10 images)
Output
image
Aspect ratios
21:9, 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16, 9:21
Cost
14 credits per run

Muse Image questions

Text-to-image starts from nothing, while edit mode starts from one to ten images you connect. Use text-to-image for posters, infographics, product shots and QR graphics that need exact text and layout; it can't lock a specific subject or face. Edit mode changes, combines or restyles existing images, but needs at least one connected image and regenerates the whole frame, since there is no mask.

Use edit mode. Connect the photo, phrase one clear instruction, and name what to keep unchanged, such as the face, pose, background and lighting. Edits can drift a little if the region is described loosely, so be specific about what to change and avoid stacking many unrelated edits in one instruction.

Up to ten. Supply a subject, a setting and a style source as separate images, and refer to them by input order in the prompt (for example @Image1 and @Image2) so the model knows which element comes from where.

Text, mostly yes. Headlines, labels and short captions come out crisp and correctly spelled most of the time; wrap exact text in quotation marks and keep strings short. Long paragraphs are weak and some runs need a re-roll. Charts and QR codes are another matter: bars are drawn by eye, not to scale, and a QR code generated from text didn't scan in our tests. Connect a real chart or code instead, and Muse copies it faithfully.

Mask-based inpainting or outpainting, long passages of body text, exact pixel dimensions, tricky mirror and reflection physics, and stylized painterly illustration. Without style cues the look is polished, neutral realism, so add style direction when you want more character.

It was created by Meta Superintelligence Labs and released on 2026-07-07. It generates and edits images in one workflow, with output reaching about 1600px on the longest edge.

Try Muse Image on Fuser

Posters, charts and QR codes with type that spells right, plus edits that change only the one thing you asked for

Render the brief

Related models

Muse Image docsAll models

Gemini Image

Google DeepMind

GPT Image

OpenAI

Seedream

ByteDance (Seed team)

FLUX.3 Image

Black Forest Labs

FLUX.2

Black Forest Labs

Topaz Image Upscale

Topaz Labs

Muse Image in our guides

All articles