Meta Unveils Muse Image AI Model

Meta Platforms rolled out its own imaging tool, Muse Image, across its AI chatbot. It is the first AI image generation model built by Meta Superintelligence Labs following the launch of the company’s initial large language model in April. Muse Image marks the first major release since Meta rebuilt its AI lab over the past year under chief AI officer Alexandr Wang.

According to Meta’s blog post, users can describe what they want in conversational language and let Meta AI handle the rest, including tasks like placing themselves in front of a landmark, removing a photobomber from a shot, or generating a functional QR code. The model also renders text cleanly within generated visuals, enabling things like legible how-to guides or infographics.

A new presets panel offers suggested prompts, letting users restore old family photos, try trending hairstyles, or turn themselves into a Claymation character or 16-bit video game hero. Users can also photograph a room and ask Meta AI to redesign it using real products sourced from the web or Facebook Marketplace.

Meta stated that Muse Image works by pairing it with the Muse Spark reasoning model, which includes planning image layout, pulling in real-time web context, and blending multiple visual references before producing a result. The tech giant is also leaning into social features. Users can tag Instagram accounts within the Meta AI app to incorporate profiles into generated images.

The company plans to expand Muse Image to more countries and additional surfaces, including Facebook, Messenger, and further areas of Instagram and WhatsApp. Advertisers and agencies will gain access in the coming weeks through Meta’s Advantage+ creative tools, which the company said will let brands adjust elements, swap styles, and generate on-brand ad variations with fewer iterations.

Basic use of Muse Image through Meta AI is free for everyday creation, with expanded usage available through Meta’s subscription plans. For safety,  all images generated with Muse Image will carry an invisible watermark, and it has built-in precautions intended to prevent the tool from being used to create content that violates its terms of service, including child sexual abuse material. Muse Image performs strongly on several benchmarks, generally outperforming Google’s Nano Banana 2 while trailing OpenAI’s ChatGPT image generator.