Meta has launched Muse Image, a new AI image generator built to create, edit and combine visuals using ordinary conversational prompts.
It is the first image generation model released by Meta Superintelligence Labs, the division formed to develop the company’s most advanced AI systems. Muse Image is now available through the Meta AI app and website, while some of its features are also beginning to appear across Instagram and WhatsApp.
Meta is pitching it as more than another tool that turns a short sentence into a picture. Muse Image can reason through complicated requests, search the web for current information, use code when precision matters and refine its own work before delivering the result.
What Can Muse Image Do?
Muse Image can generate a new visual from scratch or work with photos you already have.
Meta’s examples include:
- restoring old family photos
- removing unwanted people from backgrounds
- turning someone into a claymation character
- creating illustrations and product images
- redesigning rooms
- combining people, objects and locations from several photos
- generating infographics with readable text
- producing working QR codes
- making personalised stickers, posters and postcards
You can describe what you want in normal language rather than learning a complicated prompt formula.
For example, you could upload a room photo and ask Meta AI to give it a Japandi makeover, remove a photobomber from a holiday picture or turn a selfie into a 16-bit video game character.
Meta has also added suggested presets for users who don’t know where to begin. These provide ready-made ideas that can be applied to an existing image with one tap.
It Thinks Before Creating the Image
The more interesting part of Muse Image is what happens behind the scenes.
Most image generators try to turn the user’s prompt directly into a finished image. Meta says Muse Image instead acts more like an agent. It can plan the composition, use tools, examine its first attempt and decide whether something needs to be fixed.
If the request involves current facts or real-world details, it can search the web for additional context. If the image needs a precise chart, formula or QR code, it can write and run code to improve the result.
It can also refine its own work. That might mean making a small correction to an existing image, starting again when the whole composition is wrong or trying a different method to get a more accurate result.
Meta says the model improves when it is given more time and computing power to reason through a request, rather than simply generating several images and picking the best one.
Editing Is More Conversational
Muse Image is also designed for ongoing editing.
Instead of starting from scratch whenever something looks wrong, users can continue the same conversation and ask for changes. Meta AI remembers the earlier prompt and the image being edited, which should make it easier to swap styles, replace objects or change details over several turns.
There is also a markup tool that lets users circle, sketch or annotate the part of an image they want changed.
That could make image editing much easier for everyday users. Instead of trying to describe where an object is located, you can mark it directly and explain what you want done.
It Can Combine Several Reference Images
Muse Image can pull elements from multiple uploaded images into one composition.
That could mean placing a person from one photo into the location shown in another, combining clothing and objects from several references or creating a personalised design using different visual sources.
Meta says Muse Image can handle people, environments, styles and objects together while following written instructions placed between the image references.
This is likely to be one of its most useful features for creators, marketers and people who want to build personalised graphics without manually cutting and combining every element.
Meta Quickly Removed One Controversial Feature
Meta’s original announcement said users could @-mention public Instagram accounts and use their public photos as references when creating images.
That feature did not last long.
Meta updated the announcement on July 10, saying it had heard feedback that the feature “missed the mark” and had removed it. The company had initially said Instagram users would be able to control whether their public content could be referenced through an account setting.
The quick reversal highlights one of the difficult questions surrounding personalised AI image tools. A feature may be technically possible and still make people uncomfortable when it involves their photos, likeness or public social media content.
Muse Image can still work with photos supplied directly by the user, but the ability to reference another public Instagram account is no longer available.
Muse Image Is Moving Across Meta’s Apps
Muse Image is available through the Meta AI app and the Meta AI website.
It is also powering more than 30 new effects for Instagram Stories in the United States and image generation inside Meta AI chats on WhatsApp in a limited number of countries. Meta says broader availability is coming to additional countries and services, including Facebook and Messenger.
Advertisers and agencies will also be able to use Muse Image through Meta Advantage+ creative, giving businesses another way to produce advertising images and variations at scale.
Everyday image creation is free, although Meta says users who want to generate more can access higher limits through its subscription plans.
Meta Is Adding an Invisible AI Watermark
Images created by Muse Image through Meta AI include a hidden watermark called Content Seal.
Meta says the signal remains attached even when an image is cropped, compressed, resized or captured in a screenshot. The company is also previewing a detection tool that can check whether an image contains the watermark.
Invisible watermarking will not solve every problem around AI-generated content, but it could give platforms and users another way to identify images made through Meta’s tools.
Meta plans to extend Content Seal to generated video as well.
Muse Video Is Coming Next
Meta also previewed Muse Video, which is being developed using the same underlying training foundation as Muse Image.
The company says the video model will support native audio and is being designed for strong visual quality, prompt accuracy and consistency between frames. Meta admits it is still working on areas including fast physical movement and audio-video synchronisation.
Muse Video is expected to become available to creators and Meta AI users later.
Its development shows that Meta’s bigger plan goes well beyond still images. The company wants to build a full set of media-generation tools that sit directly inside the social apps billions of people already use.
Why this matters for Australia
Australians already spend a huge amount of time across Instagram, Facebook, Messenger and WhatsApp, which gives Meta a major advantage when introducing a new creative AI tool.
People may not need to search for a separate image generator, create another account or learn an unfamiliar platform. Muse Image can appear inside services they already use for messaging, sharing photos and creating social content.
Availability may differ between apps and countries during the rollout, so Australian users may not see every feature immediately. Meta says Muse Image will expand to more countries and more parts of its platform over time.
The bigger issue is how quickly AI image generation is becoming an ordinary social media feature. Creating and heavily editing synthetic images is moving away from specialist creative software and closer to something anyone can do inside a chat, story or feed.
Muse Image may make visual creation easier, but Meta’s quick reversal on Instagram account references shows that convenience, consent and personal likeness will remain difficult issues as these tools become more powerful.
The takeaway is simple: Meta is not treating AI images as a separate experiment. It wants them woven directly into the social platforms people use every day.
Source
Meta
