On 7 July 2026, Meta launched its first image generator, Muse Image, and put it straight inside Meta AI, WhatsApp and Instagram. It also showed off a video version called Muse Video. Then, within days, it pulled the most talked-about feature: the one that let anyone make AI images from a stranger's public Instagram photos. So there are really two stories here. A good new AI tool, and a privacy mistake Meta undid in under a week.
Straight to your question:
- What is Muse ImageMeta's answer to ChatGPT's image tool, inside apps you already use. What it does.
- And Muse VideoOnly shown, not released. Same base, plus sound, but you cannot use it yet. Why.
- The controversyIt could use other people's public photos by default. Meta undid it. The full story.
- Protect my photosThe feature is gone, but the risk is not. What to change.
- Is it any goodHow it compares to ChatGPT Images 2.0 and Google's model. See the table.
The one thing to remember: Muse Image is not a website you go to. It is already in the Meta apps on your phone. That is why the privacy worry hit so hard.
What Muse Image is, and where you use it
Muse Image is the first image tool from Meta Superintelligence Labs, the team Meta built to work on cutting-edge AI. It launched on 7 July 2026. What makes it stand out is not how good it is. It is where you use it. It did not come out as a research paper or an app you download. It came out inside Meta AI, WhatsApp and Instagram, with Facebook and Messenger next. You do not go anywhere new. Meta put an image maker into the apps you already have.
What can it do? Meta says it follows your instructions well, edits images cleanly, and can use more than one photo as a reference at the same time. Behind the scenes it works a bit like the newest tools. It can search the web for reference, use tools to work through your request, and check its own work before it shows you. In short, it thinks a little before it draws, the same change we saw when ChatGPT rebuilt its image tool.
Some of the tricks are genuinely handy. It can make working QR codes inside a picture. You can circle or sketch on an image to point at what you want changed. There is a Facebook Marketplace mode that redraws a room around a product. And it powers a batch of new Instagram Stories effects, starting in the US. One note for developers: there is no public API yet, so for now it is a feature you use, not one you can build on.
Muse Video: still just a promise
This is where the headlines got ahead of the facts. Muse Video was shown at the same time as Muse Image, so people talked about them as a pair. They are not, at least not yet. Muse Image is out. Muse Video was only shown, marked coming soon, and by late July you still cannot use it.
Here is what Meta has said. Muse Video uses the same base as Muse Image and adds sound made together with the video instead of added later. That would be a real plus, because most AI video is silent by default. But there are no test results, no length or quality numbers, and no API. So if you read that Meta launched an AI video maker, it is worth being clear: Meta launched an image maker and showed a video one. Treat everything about Muse Video as a promise until you can try it.
The feature Meta pulled within days
Muse Image made the mainstream news for one reason, and it was not the pictures. It was a single choice. At launch, one way to make an image in Meta AI was to @-mention a public Instagram account, and the tool would create pictures based on that person’s photos, style or looks. The catch: this was switched on by default for public accounts, you had to turn it off yourself, and the person was never told.
People reacted fast, from privacy groups to public figures like the Emmy-winning actor Hannah Einbinder. The problem was easy to see: anyone could make changed images of you, from your own photos, without you knowing or agreeing. Within days Meta backed down. Its statement was unusually plain. The idea had been to give people a useful creative tool and let them control whether their public content could be used this way, but it had heard the feedback that the feature missed the mark, so it was no longer available.
The lesson is simple. The model was fine. The default setting was the problem. Turning on a feature like that by default, inside the biggest photo app in the world, was always going to end badly. With AI features, the real question is rarely can it do this. It is who gets switched on without being asked.
How to stop AI tools using your photos
The @-mention feature is gone, but the worry behind it is not, and it is bigger than Meta. A few simple steps, none of them slow:
- Make your account private if it does not need to be public. Nearly every “my photos were used for AI” story starts with a public profile. On Instagram: Settings, then Account privacy. A private account cannot be used by strangers at all.
- Check Meta’s AI settings. In the Instagram or Facebook Accounts Centre, look in the privacy and AI sections for controls over how your content feeds AI, and turn off what you do not want.
- Think before you post a face that is not yours. Same rule as ChatGPT: your own photos are your choice, other people’s are not, and children’s faces should stay out of any AI tool.
- Watch the defaults after every AI update. The Muse lesson is that new features can arrive already switched on. When an app adds AI, spend two minutes in the settings instead of trusting the default.
If privacy is your main worry with AI image tools in general, our guide to ChatGPT Images 2.0 covers the same settings question on OpenAI’s side.
Muse Image vs ChatGPT Images 2.0 and Google
Muse Image did not arrive alone. OpenAI and Google were already going at it, and the early word from people running side-by-side tests is that this is now a real three-way race. Here is the shape of it, as of late July 2026.
| Muse Image | ChatGPT Images 2.0 | Google (Nano Banana 2) | |
|---|---|---|---|
| Made by | Meta Superintelligence Labs | OpenAI | |
| Where you use it | Instagram, WhatsApp, Meta AI | ChatGPT (web and app) | Gemini app and Google apps |
| Best at | Being inside social apps; guided editing | Following instructions; text in images | Fast photo edits by chat |
| API for developers | Not yet | Yes (gpt-image-2) | Yes |
| Video version | Muse Video (preview) | Sora | Veo |
The point is less about who wins on pure image quality, and more about where you already are. If you live in Instagram and WhatsApp, Muse Image is the one you will reach for, because it is right there. If you want the best instruction-following, or you want to build it into your own app, ChatGPT Images 2.0 and its API are ahead. If you use Google’s apps, its tool is there for you. For most people, the best AI image tool in 2026 is just the one inside the app they open most. Meta knows that, and it is the whole reason it put Muse Image where it did.
One early snag, from users rather than Meta: some say the image-to-video and animation steps work in old chats but fail in new ones, the kind of rough edge you get with a brand-new feature. If it fails, try again in a fresh chat, or wait a day.
Should you use it?
If you are curious, go ahead. It is free and already in your Meta apps. It is good at guided edits and at the social tricks like Stories effects and QR art. Just keep two things in mind. First, check your privacy settings before you start, for the reasons above. Second, keep your hopes realistic. Muse Image is a strong new option, not a clear winner, and Muse Video is not really here yet, whatever the headlines said.
For serious work, where you need an API, the best instruction-following, or the same result every time, the older tools still have the edge for now. But as a free, built-in tool that will only get better, Muse Image is worth ten minutes of your time.
Common questions
Muse Image is the first AI image tool from Meta Superintelligence Labs, launched on 7 July 2026. It makes and edits images from text, and it lives inside Meta AI, WhatsApp and Instagram rather than as a separate website or app.
Muse Video is Meta's AI video tool, shown at the same time as Muse Image. It uses the same base and adds sound made with the video. But it was only previewed, not released, so as of late July 2026 you cannot use it yet.
Yes. It is built into Meta's apps, so if you have Instagram, WhatsApp or Meta AI you already have it at no extra cost, within the usual limits Meta sets.
At launch, people could @-mention any public Instagram account to make AI images based on that person's photos, and this was on by default without telling them. After a strong privacy backlash Meta undid it within days, saying the feature had missed the mark and was no longer available.
The feature that let strangers do this has been removed. To lower the wider risk, set your account to private, check the AI and privacy controls in Meta's Accounts Centre, and remember that a private account cannot be used by strangers in the first place.
Make your profile private, then check the privacy and AI sections of the Instagram or Facebook Accounts Centre for controls over how your content feeds AI, and switch off anything you do not want. Doing this after each big app update is the safest habit.
It depends on what you want. Muse Image's advantage is being built into apps you already use, plus good guided editing. ChatGPT Images 2.0 leads on following instructions and has a developer API. For most people the better tool is just the one inside the app they use most.
Not yet. At launch it is a feature inside Meta's apps with no public API, unlike OpenAI's gpt-image-2. Meta may add one later, but there is nothing to build against today.
It makes and edits images from text, works from several reference photos at once, can make working QR codes, lets you circle or sketch to edit, redraws rooms around Marketplace products, and powers new Instagram Stories effects. It also uses web search and checks its own work to improve results.
It is the team Meta set up to work on cutting-edge AI, and Muse Image and Muse Video are its first image and video tools. The launch is an early sign of how the team works: putting AI straight into Meta's apps rather than releasing it as research.
Sources
- Launch and models: Meta AI, Introducing Muse Image and Muse Video; Meta Newsroom.
- Privacy backlash and reversal: The Guardian, Meta ditches Muse Image feature; Meta’s rollback statement.
- What shipped vs previewed: Muse Image and Muse Video deep dive.
