Muse Video (not available)
Muse Video was announced as an early preview alongside Muse Image. Meta’s public post says it is coming soon to creators and Meta AI. It is not in Meta Model API.
| Path | Status |
|---|---|
| Path A — first-party API | No video_generation (or equivalent) endpoint. Model API video APIs are understanding only (Muse Spark reads video → text). |
| Path B — local / open weights | No Hugging Face / Diffusers checkpoint. Muse Glimmer is open-weight but outputs text, not video. |
| Third-party hosters | Not wired. OpenTryOn only adds Fal/Replicate/etc. when you explicitly ask. |
There is no --model muse-video (or local twin) in the registry. Do not confuse Muse Spark video understanding with video generation.
When Meta ships a first-party generate/edit video API or open weights, follow Model Integration Guidelines the same way as MiniMax H3: two ModelSpec ids if both paths exist, never one adapter that mixes HTTP and GPU.
Until then, use existing video models (sora, ltx-2.5-api, minimax-h3, wan-3.0, …). Image twin: Muse Image.