Meta Platforms launches Muse Image, a new AI image-generation model and its first release from its Meta Superintelligence Labs. The model is integrated into Meta AI and is made available across Meta AI, Instagram Stories in the U.S., and WhatsApp in limited countries, with additional placements such as Facebook described as coming later. In addition to generating images from prompts, Meta says Muse Image supports editing and refinement across multiple turns, including composing images using multiple reference photos and blending elements such as people, objects, styles, and environments. Meta also previews Muse Video, a separate model built on the same pretraining base, with availability for creators and Meta AI planned but not yet for general use.
Multiple outlets report that Muse Image can draw from public Instagram photos when users include public Instagram accounts in their prompts, while users can opt out via Instagram settings. Meta also includes a provenance feature called Content Seal, described as an invisible watermark that persists through common image transformations, and says it plans to extend it to video and offer a detection tool. Meta says the system is designed to follow instructions and improve factuality, and some reports describe a tool-using, self-refining approach. The rollout follows Meta’s rebuilding of its AI lab over the past year.
