Meta ships an agentic image model and previews native-audio video
Meta launched Muse Image in Meta AI, meta.ai and selected messaging and Stories surfaces. Instead of relying on one diffusion pass, the system can search for references, write code for precise layouts and repeatedly inspect and refine its output. Meta also previewed Muse Video, a text-to-video model designed to generate synchronized audio natively.
Why it matters: Image generation is shifting from a single model call toward a composed workflow that can retrieve evidence, calculate layouts and critique its own drafts.