← Back to articles
How to Generate and Edit AI Images with Meta's Muse Image

How to Generate and Edit AI Images with Meta's Muse Image

You’re texting a friend on WhatsApp about a birthday party, and instead of scrolling stock photos for an invite, you just describe the one you want and get it in seconds, in the same chat. No new app, no account to create, no waiting for a design to come back from someone else. That’s the shift Meta just made with its newest AI image tool, and it’s already sitting on your phone whether you’ve noticed it or not.

This week’s biggest practical AI story is Muse Image, Meta’s new image generation and editing model built directly into the Meta AI experience on WhatsApp and Instagram. It’s free, it’s already rolling out, and it works differently from most image generators you may have tried before. Here’s what it actually is and how to start using it.

What Is Muse Image, Really?

Most people assume an AI image generator just turns your words into a picture and stops there — you type a prompt, get an image, and start over if it’s wrong. That’s roughly how earlier tools worked, and it’s still how a lot of people think about “AI art.”

Muse Image works differently. It’s built as an agent, not a one-shot generator: instead of mapping your prompt straight to pixels, it can pause, check its own work, and pull in outside help — like running a quick search or a bit of code — to make sure text, logos, or facts in the image are actually correct. It also accepts more than just typed prompts. You can upload an existing photo and edit it, or sketch directly on an image to point at exactly what should change, and Muse Image remembers the conversation so you can keep refining instead of starting over every time.

The other big difference is where it lives. You don’t need to visit a separate website or download a new app — Muse Image runs inside the Meta AI chat you already have access to on WhatsApp and Instagram, which is part of why it’s spreading so fast.

How Does It Work?

Think of Muse Image less like a vending machine (put in a prompt, get out one image) and more like a very fast junior designer sitting in the chat with you. You describe what you want, they sketch a first draft, and instead of just guessing, they’ll double check spelling on a sign or the shape of a logo before handing it back. If you’re not happy, you don’t need to write a whole new request — you can circle the part that’s wrong and say “fix this,” and they’ll remember what you were making in the first place.

Under the hood, this works because the model spends extra processing time while generating a single image — checking and refining before handing it back, rather than committing to one pass. Combined with its ability to call outside tools like search or code, that’s what lets it hit small details (accurate text, working QR codes, consistent branding) that older single-pass generators regularly got wrong.

How to Try It Yourself

Muse Image is free to use today inside Meta AI on WhatsApp and Instagram in the US, with more surfaces like Facebook and Messenger coming soon.

  1. Open WhatsApp or Instagram on your phone.
  2. Start or open a chat with Meta AI (look for it in your chat list, or search for it if you don’t see it).
  3. Type a plain-language description of the image you want, e.g., “a cozy birthday invite with balloons and the words ‘Turning 30!’”
  4. Wait for the image to generate, then look for a markup or pencil icon on the result.
  5. Tap it to circle or sketch directly on the part you want changed, and describe the fix (e.g., “make the balloons blue instead of red”).
  6. Keep refining in the same conversation — Muse Image remembers what you’re working on, so you don’t need to re-describe the whole image each time.
  7. Save or share the final image directly from the chat once you’re happy with it.

If you also want to combine details from more than one photo — like putting a specific outfit on a specific background — you can upload multiple reference images in the same conversation and describe how you want them combined.

Tips to Get Better Results

Describe the scene, not just the subject. “A dog” gives Muse Image very little to work with. “A golden retriever sitting on a porch at sunset, warm lighting” gives it an actual scene to build.

Use the sketch tool for precision edits. If a generic instruction like “fix the sign” isn’t landing, circling the exact spot and describing the fix gets you there faster than rewriting the whole prompt.

Lean on it for text-heavy images. Because Muse Image checks its own work, it’s noticeably better than older tools at getting things like invitation text, signage, or QR codes to render correctly — a good use case if past AI images left you with garbled lettering.

Stay in one conversation while you iterate. Starting a fresh chat means losing the context Muse Image built up about your image. Keep refining in the same thread until you’re done.

Double-check anything meant to be factual. The model’s self-checking helps, but it’s still worth a quick human glance at logos, maps, or diagrams before you send them somewhere that matters.

Closing Thought

You already have this tool available if you use WhatsApp or Instagram — there’s no signup, no waitlist, no new app to learn. Open a chat with Meta AI right now and describe one image you actually need this week, whether it’s a party invite, a thumbnail, or just something fun. The best way to understand what an agentic image tool can do differently is to hand it a real task and watch how it handles the back-and-forth.