Meta launches Muse Image, its first Superintelligence Labs model, and previews Muse Video

Meta released Muse Image on July 7, the first publicly shipped model from Meta Superintelligence Labs, rolling it out inside the Meta AI app, Instagram Stories in the US, and WhatsApp in limited countries. Meta also previewed a companion video model, Muse Video, still in early development.

What's new

Muse Image is built as an agentic image system rather than a straight prompt-to-image model. According to Meta's announcement, "Muse Image operates as an agent: it invokes search and coding tools to improve accuracy, self-refines its own generations, and improves through scaling test-time compute." In practice that means the model can call a web search tool to ground factual details (useful for infographics or current-events imagery), use coding tools to render precise elements like charts or QR codes, and iteratively critique and refine its own output across multiple reasoning passes before returning a final image.

Other capabilities Meta is emphasizing:

• Multi-reference composition — blending several input photos into one new image
• Precise, turn-based editing — changing only the specific elements a user asks for, across repeated edit rounds
• Instagram social context — drawing on a user's Instagram activity to inform stylistic choices
• Content Seal — an invisible watermarking layer for provenance verification on every generated image

Muse Video, meanwhile, is described as an early preview built on the same pretraining base as Muse Image, with Meta saying it "offers competitive performance in prompt adherence, visual fidelity, and temporal consistency," and that it includes native audio generation. No release date was given.

On availability, Muse Image is live now in the Meta AI app and on meta.ai, in Instagram Stories (US only), and in WhatsApp in a limited set of countries. Meta says Facebook support and broader Muse Video access are coming later. Meta AI's Muse Image use is free for everyday creation, with expanded usage folded into Meta's existing subscription plans; the model will also power advertiser image tools inside Meta's Advantage+ creative suite.

Context

Muse Image is the first model to ship publicly from Meta Superintelligence Labs, the unit Meta built out under Alexandr Wang after its 2025 reorganization of AI efforts, following the unit's earlier Muse Spark foundation model release in April. It puts Meta into direct competition with image tools from OpenAI, Google, Midjourney, and xAI, all of which have shipped agentic or multi-turn editing features into their consumer chat products over the past year.

Why it matters

The agentic framing — search-grounded generation, self-refinement loops, test-time compute scaling — mirrors the direction rivals have taken with their own image tools, suggesting this has become the baseline architecture for frontier image generation rather than a differentiator. For Meta, shipping through Instagram and WhatsApp gives Muse Image an enormous built-in distribution advantage over standalone tools, and the advertiser integration through Advantage+ signals Meta intends to monetize the model directly through its ad business rather than treating it purely as a consumer feature. Muse Video's mention of native audio, if it ships as described, would put Meta in more direct competition with Runway, Luma, and Google's Veo line on end-to-end audio-video generation.