Meta has rolled out Muse Image, the first image-generation model from Meta Superintelligence Labs. It is live now in Meta AI, where it works alongside Muse Spark, the reasoning model the lab shipped in April. The pitch is straightforward: describe what you want in plain language, attach one or several photos, and the model handles the rest.
The interesting part is what happens before the image gets made. Muse Image pairs with Muse Spark to reason through a prompt first. It plans the layout, pulls in real-time web context when needed, and can blend multiple photo references into a single result. In practice, that means you can combine a selfie with a vacation photo into a postcard, or drop your dog into a famous painting, without writing anything more elaborate than the idea itself.
What it does
Beyond generation from scratch, the model handles photo restoration, object removal, and legible text within images, which opens the door to applications such as infographics, invitations, and functional QR codes. There is a presets panel with suggested prompts for people who freeze at a blank input field. One tap restores an old family photo or turns you into a claymation figure or a 16-bit game character.
Editing works directly on the image. Tap the markup icon, circle or sketch what should change, and refine from there. Since Meta AI retains the conversation context, you can keep adjusting styles and details without having to start over.
One feature that leans on Meta’s home turf: photograph your room and ask for a redesign using real products from the web or Facebook Marketplace, styled to a trend or a look you name yourself.
Where it lives
Muse Image already powers over 30 new AI effects in Instagram Stories and image generation in Meta AI chats on WhatsApp, with a limited rollout in select countries. Facebook and Messenger are next, and advertisers will get access through Advantage+ creative in the coming weeks. Everyday use is free, with heavier usage included in Meta’s subscription plans.
A quick retreat on Instagram mentions
The launch came with one stumble. At release, users could generate images by @-mentioning public Instagram accounts as visual references. The feedback was swift enough that Meta pulled the feature within three days, acknowledging it “missed the mark.” A reminder that even opt-in-by-default framing around other people’s faces tends to land badly.
What comes next
Meta says images are only the first step. Muse Video is already in development, part of what the company frames as its path toward personal superintelligence. For now, Muse Image puts Meta back in direct competition with the image tools from OpenAI and Google, with one structural advantage the others lack: distribution across apps that a few billion people already open every day.