Meta has launched Muse Image, marking the company’s first image generation model from Meta Superintelligence Labs.
The model operates across Meta AI chatbot, Instagram Stories, and WhatsApp, replacing Midjourney as Meta’s image generation provider. Alexandr Wang, chief AI officer and head of Meta Superintelligence Labs, led the development effort after Meta hired him in 2025 to rebuild the company’s AI capabilities.
Muse Image generates images from conversational text prompts without requiring precise prompt engineering. Users describe scenes in natural language and the model translates those descriptions into images. The model also edits existing photographs by removing elements, adjusting camera angles, and modifying lighting and composition. Meta AI provides a sketch tool that allows users to hand-draw edits and have the model refine them accordingly.
“This initial model is small and fast by design, yet capable enough to reason through complex questions in science, math and health. It is a powerful foundation, and the next generation is already in development,” the company said in a blogpost, revealed two months ago.
The model integrates with Muse Spark, Meta’s reasoning language model released in April. Muse Image invokes Muse Spark to reason through complex requests before generating images. The combined system plans layouts, searches the web for real-time context, and blends multiple visual references simultaneously. Users can place themselves paragliding over exotic landscapes, combine selfies with vacation photos to create postcards, or drop pets into famous paintings.
Muse Image features agentic tool use that allows it to invoke coding and search capabilities. The model writes Python code to generate accurate plots and QR codes. Web search integration grounds generated images in factual and real-time information. The model self-refines outputs by analyzing its own work and correcting errors before displaying results to users.
Meta benchmarked Muse Image against competitors and reported it surpasses Google’s Nano Banana 2 on several evaluation metrics while trailing only OpenAI’s latest image generator in overall quality. Instagram Stories received over 30 new AI-powered effects powered by Muse Image. WhatsApp users in select countries can generate images directly in chats with Meta AI.
A notable feature enables users to tag Instagram accounts by username, pulling public photos from those profiles to incorporate into generated images. Users can disable this feature in settings to prevent their photos from appearing in AI-generated content. Meta plans to expand Muse Image to Facebook and Messenger in coming weeks and make it available to advertisers through Advantage+ creative tools.
