← Back to Video models
Meta AINotable
Muse Video
Released Jul 7, 2026
Meta Superintelligence Labs' video generation model with native audio, built on the same pretraining base as Muse Image and shown as an early preview.
- Price per second
- —
- Max clip length
- —
- Max resolution
- —
- Native audio
- Yes
- AA arena Elo
- —
- LMArena Elo
- 1456
- Capabilities
- text to video
- Licence
- Proprietary
Strengths & weaknesses
Strengths
- Native audio support
- Meta cites competitive prompt adherence, visual fidelity and temporal consistency
- Meta reported a No. 3 text-to-video Elo on Arena at announcement
Weaknesses
- Early preview only: 'coming soon to creators and in Meta AI', no public access or API
- Meta names gaps in audio-video synchronization and physically accurate fast motion
releaseDate is the preview announcement (Meta AI blog, July 7 2026); the model is not generally available. Meta's developer models page (dev.meta.ai/models) lists Muse Image but no Muse Video, so no API id exists. Only text-to-video is listed because Meta cites a text-to-video Arena rank; image-to-video, editing, duration and resolution are not stated. openWeights false is inferred from the absence of any release of weights (no HF/GitHub page found), not from an explicit statement. Not on Artificial Analysis' board. LMArena name 'muse-video' (#9, Sep 21 2026).