Meta Launches New AI Tool That Can Generate Images From Text Prompts

Meta's Superintelligence Labs has released Muse Image, an AI tool that creates pictures based on text descriptions. You can use it right now through Meta AI, Instagram, and WhatsApp, with Facebook and Messenger support coming soon The Verge.

Here's how it works: you type a description of what you want to see — say, "a sunset over mountains" — and the AI generates an image based on your words. What makes Muse Image different from similar tools is that it doesn't go straight from your words to an image. Instead, it first thinks through your request with another AI called Muse Spark, searches the web for ideas, and plans before creating the picture. That's different from older AI image tools, which just converted text directly into pictures in a single step The Verge Threads.

What You Can Actually Do With It

The tool has several features built into Meta's social platforms. You can mention someone's Instagram account inside your prompt — like "@friends_account" — and the AI will use their public photos to include their face or style in the generated image. Meta says it pulls from that person's public posts to build the likeness. If you don't want your photos used this way, you can turn it off in Instagram's settings Instagram Help.

Beyond that, you can transform images by writing new descriptions, design invitations and postcards, redesign a room based on a photo, or draw directly on a photo and then share the edited version on Instagram. Instagram Stories will also get 30 new AI effects powered by Muse Image, arriving first in the United States The Verge Instagram.

Meta also announced that a Muse Video model — which would generate videos from text — is in the works, but hasn't said when it will arrive Meta AI Blog.

Where Muse Image Fits In Meta's Larger Bet

Muse Image is the second product released by Superintelligence Labs, the new division Meta created after paying $14 billion to bring in leadership to run its AI efforts. The first product was Muse Spark, a general-purpose AI assistant that now powers Meta AI across WhatsApp, Instagram, Facebook, and Messenger Meta The New York Times. That $14 billion deal was one of the largest single investments in AI talent in the industry CNBC.

However, while Meta rolled out the consumer-facing version quickly, it has taken a slower approach with developers. As of early June 2026, Meta had delayed releasing Muse Spark's developer API — the tool that would let outside programmers build on top of it — multiple times with no firm release date announced Reuters. Meta hasn't said whether it will open up Muse Image to outside developers on a different timeline, or at all Meta.

The Privacy Question

The feature that will likely draw the most scrutiny is the one that lets you tag someone's Instagram account to incorporate their likeness. This matters because it turns a library of photos that people posted for social sharing into material that feeds the AI's image generation — all tied to a single person.

Meta's approach here deserves attention. The company is making this opt-out rather than opt-in, which means your photos are used by default unless you go into Instagram's settings to turn it off. You have to actively choose not to participate. This follows a familiar pattern from Meta's past: the platform gets the benefit of the feature, and individual users have to remember to restrict it. There is already scrutiny of this kind of data use coming from regulators and privacy advocates, and whether this default will hold up under closer examination remains an open question.

Separately, the Trump administration has been pressing Meta to allow its AI models to be reviewed by the government for security concerns Reuters. Meta hasn't said whether Muse Image or Muse Spark will be covered by any such review.

The Broader Picture

What's emerging is a pattern: Meta is moving quickly to get consumer features into the hands of its billions of users, while proceeding much more carefully when it comes to handing tools to outside developers. That gap is normal for a company trying to manage risk, but it's worth noting given how much Meta spent to get back into the race for cutting-edge AI.

The next real test will be whether a video generation model arrives with the same consumer-first approach, and whether the kind of thinking Muse Image does — searching the web and planning before generating — holds up once it encounters the volume and creativity of Instagram's entire user base.