Meta announced the launch of Muse Image and previewed Muse Video, its first media generation models developed by Meta Superintelligence Labs. Muse Image is described as the company’s most advanced image generation model, capable of following instructions, editing with precision, and composing from multiple references. It also integrates with Muse Spark and supports agentic tool use. Muse Video, built on the same pretraining base, delivers exceptional visual fidelity with native audio support and is set to release soon for creators and Meta AI users. Source: metaai
Muse Image operates as an agent, using search and coding tools to improve accuracy and self-refine its generations. It integrates with Muse Spark to enable joint planning for agentic media generation. The model can also use coding tools during reinforcement learning to produce accurate plots and QR codes, and it learns to search the web for factual and real-time information to ground its generated images. Search tool use improves the model’s factual accuracy on knowledge-intensive prompts, particularly those involving current events and real-world facts. Source: metaai
Muse Image maintains coherence across editing turns, supporting iterative refinement and open-ended brainstorming toward a target result. It can compose elements from multiple input reference images, including people, objects, clothing, styles, and environments, and supports interleaving text and images in prompts for complex compositions. The model holds the No. 2 spot on Arena for text-to-image, single-image editing, and multi-image editing as measured by human preference Elo rankings as of July 5, 2026. Source: metaai
Source: metaai
