Muse Image

Muse Image plans your shot, searches the web for context, and self-refines drafts before you see them. See real examples and how to try it today.

Muse Image doesn't just turn a prompt into a picture. As a muse AI image generator, it plans, searches the web for context, blends your reference photos, and refines its own draft before you ever see it — all inside Meta AI.

Enter a prompt, add up to 10 reference images, and create your own image directly on this page.

Optional reference images ( 0 / 10 , JPG/PNG up to 10MB each) — leave empty to generate from text only.

Muse Image is Meta's agentic AI image generator, built by Meta Superintelligence Labs. Instead of mapping your prompt straight to a picture, this muse image generator plans ahead, calls on search and coding tools, and self-refines its own draft to get closer to what you actually asked for. Muse Image also integrates with Muse Spark, so the two models can share tools and plan jointly for more advanced, agentic media generation.

Muse Image reasons through a prompt in steps and uses tools along the way, so you get a more accurate result than a single-pass text-to-image model.

The model reviews its own draft and self-refines it — a small local fix or a full redo — before you ever see the result.

Like a reasoning language model, Muse Image improves the more compute it spends at inference time, so quality scales with reasoning depth, not just model size.

From agentic tool use to precise editing, here's what powers