Robots Atlas>ROBOTS ATLAS
Muse Image
AI Models›Muse

Muse Image

1.0 · Family: Muse
The first proprietary image generation model from Meta Superintelligence Labs. Launched July 7, 2026 in Meta AI, Instagram, and WhatsApp. Replaced Midjourney inside Meta AI.
✓ Active✓ Public access★ FeaturedImage generationMultimodal📁 Muse
Context window
Nie dotyczy (model image gen; wielo-obrazowe blendowanie zamiast okna tokenowego)
tokens
Parameters
Nieujawnione publicznie
parameters
Release date
7 July 2026
🏢Meta AIProducer
Access:HostedDeployment:☁ Cloud

Overview

Muse Image is the first proprietary image generation model from Meta Superintelligence Labs (MSL), Meta's AI unit led by Alexandr Wang. Announced on July 7, 2026, it is available in Meta AI, on Instagram and WhatsApp, and is coming soon to Facebook and Messenger. Internal project codename: Mango.

Replacing Midjourney in Meta AI

Until now, Meta AI relied on Midjourney's technology for image generation. Muse Image is a strategic move to Meta's own stack: full control over the model, integration with Meta apps, and the end-user experience. This is the first real test of the new MSL structure, established on June 30, 2025 in Menlo Park to unify Meta's dispersed AI teams.

Reasoning instead of pure generation

Muse Image does not create an image in a single pass — it first analyses the prompt together with Muse Spark (MSL's sibling reasoning model). The behind-the-scenes pipeline: planning composition, looking up real-time web context, and selecting and intelligently blending multiple visual references. Thanks to this, the model handles complex instructions like 'place my pet in a famous painting' or 'combine a selfie with a vacation photo into a personalised postcard'.

Key capabilities

• Text rendering in images — legible, correctly spelled text inside graphics. You can request a how-to guide or a detailed infographic and the text comes out legible and styled to match the composition.

• Multi-image blending — the model combines multiple input photos into one coherent scene while preserving people's identity.

• Prompt presets — a panel of suggested prompts (restoring an old photo, turning you into a 16-bit game character, claymation style, Renaissance master, and more).

• Direct on-image editing — the user circles, sketches or annotates directly on top of the generated image; Meta AI remembers the full context of the conversation, so further tweaks do not require starting from scratch.

• Room redesign — photo of a room + prompt = restyling with real products from the web or Facebook Marketplace.

Ecosystem: Meta AI, Instagram, WhatsApp, Advantage+

Muse Image also powers over 30 new AI effects in Instagram Stories and image generation in chats with Meta AI on WhatsApp. Advertisers and agencies will get access through Meta Advantage+ creative. Basic use of Meta AI with Muse Image is free; for intensive creators — as part of Meta's subscription plans.

Strategic significance

Muse Image is MSL's first media model and a signal that Meta is back in the game after falling behind in the AI race in spring 2025. Together with Muse Spark (LLM announced in April 2026) and the announced Muse Video, Meta is building its own family of foundation models replacing Llama and competing with OpenAI (DALL·E, Sora), Google (Imagen, Veo), and studios such as Midjourney or Black Forest Labs.

Classification
Image generationMultimodal
Family: Muse
Applications
Image generationImage editingCreative advertisingInterior design visualizationSocial media contentPersonal visual expressionContent generationCreative writing
Access & deployment
Hosted
Cloud
Weights: Closed
Key parameters
📏 Context: Nie dotyczy (model image gen; wielo-obrazowe blendowanie zamiast okna tokenowego)
🧩 Parameters: Nieujawnione publicznie
📥 Input: text, image

Technical specification

Context window
Nie dotyczy (model image gen; wielo-obrazowe blendowanie zamiast okna tokenowego)
tokens
Parameters
Nieujawnione publicznie
parameters
License
Proprietary (własność Meta Platforms; użycie zgodnie z warunkami Meta AI oraz Meta subscription)
Hardware requirements
Not applicable — model available only as hosted in Meta's cloud (no local deployment). No publicly disclosed hardware requirements for an on-device variant.
Modalities
⬇ Input
textimage
⬆ Output
image

Capabilities and applications

Native model capabilities
Text-to-image generation
Generating an image from a text description (prompt). The model interprets a natural-language instruction and produces a new, coherent visual from scratch — without any input image.
Category: vision
Image editing
Modifying an existing image based on a text instruction or direct annotations: removing objects, changing style, adding elements, filling in regions (inpainting/outpainting), while preserving the identity of people and scene coherence.
Category: vision
Multi-image blending
Intelligently combining several input images into one coherent composition — e.g. placing a person from one photo into a scene from another, composing characters from different sources, or building collages that preserve style.
Category: vision
Text rendering in images
Generating images containing legible, correctly spelled text — infographics, posters, menu cards, QR codes, captions in a specific graphic style. A key capability that separates new-generation models from early image generators.
Category: vision
Reasoning-augmented image generation
Multi-step generation: the model first analyses the prompt (planning layout, looking up web context, choosing references) and only then produces the image. This lets it accurately reflect complex, ambiguous instructions.
Category: reasoning
Reference-guided generation
Creating images based on previously supplied visual references — a specific person, artistic style, product, or space — preserving likeness and characteristic features.
Category: vision
Multimodal understanding
Category: multimodal
Interleaved Multimodal Input
Category: reasoning
Application domains
Image generationImage editingCreative advertisingInterior design visualizationSocial media contentPersonal visual expressionContent generationCreative writing

Technical architecture

Core Architecture
NMNative Multimodal
Model Form
MLMultimodal LLM

Sources and related pages

8 sources
BlogMeta Newsroom — Introducing Muse Image: Image Generation Built for Your World (2026-07-07)BlogAI at Meta — Introducing Muse Image and Muse Video (MSL)BlogCNBC — Meta debuts Muse Image, Superintelligence Labs' first AI image model (2026-07-07)BlogTechCrunch — Meta rolls out Muse, a new AI image generator (2026-07-07)BlogReuters — Meta expands generative AI tools with Muse Image rollout (2026-07-07)BlogThe New York Times — Meta Unveils an A.I. Image Generator (2026-07-07)BlogDataCamp — Muse Image: Meta's New AI Image Model ExplainedBlogStudioGlobal — Meta Muse Image: Inside the First In-House AI Image Generator From Meta
Browse related topics
📁 Muse🌐 Image generation🌐 Image editing🌐 Creative advertising🌐 Interior design visualization🧠 Native Multimodal🧠 Multimodal LLMAll image generation model modelsAll multimodal model models