Overview
- Meta announced Muse Image on July 7 and began rolling it out inside the Meta AI app with integrations into Instagram and WhatsApp.
- Muse Image uses an agentic multi-step process that plans work, calls web tools and computations, iterates outputs, and can render precise text and usable QR codes.
- Users can @mention public Instagram accounts to pull photos into creations and can opt out of that use, and all AI outputs carry an invisible Content Seal watermark for provenance.
- Basic generation is free inside Meta AI up to a quota and heavy use requires a paid subscription, and Meta plans to open model access to advertisers through Advantage+ creative and Marketplace product flows.
- Early Arena benchmark results place Muse Image and Muse Video near the top of image and text-to-video leaderboards, positioning Meta as a stronger competitor to OpenAI and Google as the wider rollout continues.