Muse Video

Muse Video is a text-to-video model from Meta Superintelligence Labs, previewed as part of Meta's broader agentic media stack that also includes Muse Image, Muse Spark, and Meta AI. The model shares its pretraining base with Muse Image, which the team says helps keep visual quality consistent across image and video generation. Native audio is generated in the same pass as the visuals, and the model was reported to rank among the top three text-to-video models by human-preference Elo in its launch preview.

The website at musevideo.dev is an independent, third-party platform — it is expressly not affiliated with or endorsed by Meta Platforms, Inc. On that site you get a browser-based generator: describe your scene with subject, camera movement, lighting, dialogue, and audio cues, add reference images, then pick resolution, aspect ratio, and clip length and generate. The same platform rounds out the experience with Image-to-Video, Reference-to-Video, video edit, and video extend tools, plus a model picker listing competing systems such as Seedance, Kling, and Wan, so creators can benchmark Muse Video's output against alternatives before committing.

The tool is aimed at social creators, marketers, filmmakers, and brand teams who want Meta-grade visual fidelity in a browser without a production pipeline. Generation is credit-based, and because the model is still in preview, the site notes that video generation can occasionally be temporarily unavailable — in which case no credits are deducted.