Meta's Muse Image launches free in Meta AI
Meta Superintelligence Labs released Muse Image, ranking second in human-preference image tests and free to use in Meta AI, Instagram Stories, and WhatsApp.
Meta Superintelligence Labs released Muse Image on July 7, its first image generation model. A companion model, Muse Video, was announced at the same time but listed as coming soon.
Muse Image is available now in the Meta AI app and on meta.ai, rolling out to Instagram Stories for US users, and launching in WhatsApp in limited countries. Facebook and additional Instagram surfaces follow in coming weeks. Basic image creation is free through Meta AI. Expanded creation is part of Meta’s paid subscription plans.
How it ranks
As of July 5, Muse Image holds second place in three Arena categories: text-to-image generation, single-image editing, and multi-image editing. Arena rankings are based on human-preference Elo scores across head-to-head comparisons, not automated benchmarks. Muse Video holds third place in text-to-video on Arena.
What it can do
The model accepts complex prompts referencing multiple source photos at once (multi-photo blending), renders text inside generated images, restores and edits existing photos, applies style transformations (artistic, historical, animated, claymation), and supports sketch and markup-based direct editing. Instagram users can @mention public accounts to incorporate profile photos into a generated image.
Meta embeds Content Seal invisible watermarking in every output for provenance tracking. The model integrates search and coding tools during generation and exhibits emergent self-refinement: quality improves with additional test-time compute along a log-linear relationship.
What this means for working readers
Meta’s distribution advantage here is real. Muse Image launches inside apps that already have billions of users, which means no signup friction and no separate tool to test. The @mention feature for Instagram profile photos is the most distinctive differentiator; no standalone image tool offers an equivalent.
The second-place Arena ranking across both creation and editing tasks is meaningful because it is human-preference based and covers a broader capability range than a single benchmark. The gap between first and second on Arena changes constantly, so the ranking is best read as “within striking distance of the field leader” rather than a fixed statement.
For business use, verify licensing terms before using generated images in client-facing work. The guide on AI images and commercial licensing covers what to check before images ship. Meta’s Content Seal watermark simplifies provenance tracking, but downstream terms still apply. The Midjourney AI disclosure brief covered how the image gen field has been converging on disclosure standards; Meta’s watermark approach follows that trajectory.