Meta Muse Video: The Reference Page for Meta's Previewed Video Model

Meta Muse Video does not have a release date. That is the whole answer to the question most people arrive with, and it is the answer Meta's own pages give: the model was shared as an early preview, it is described as coming soon to creators and Meta AI, and no general availability date, price, resolution or clip length has been published since. This page is the reference destination for the name — what Meta Muse Video is, what Meta claims for it, where it sits on a public leaderboard, and what you can and cannot do with it today. It is not a second announcement: the family's announcements already exist, and one of them is our own coverage of the preview and the API leak around it, which this page links to once and does not restate.

Before the substance, the reasoning, because a reader deserves to know what kind of page this is. We are past the seven-day window that normally decides whether a model is worth writing about — this page was written on 2026-09-29, the window opened on 2026-09-22, and Meta's preview of Meta Muse Video predates that window. The exception we are standing on is not a news exemption. This is a canonical reference page for a model that is live in our catalogue of pages today: the destination a reader reaches after searching the model's own name rather than the name of an event. Its warrant is standing search demand that we measured first-party on 2026-09-29, not the freshness of a launch — over the 28 days to 2026-09-25, our pages about this family earned 2,530 impressions across eighteen URLs, and every one of them hangs off a single page about a leak. Readers are asking the question and finding the wrong page. The datable anchor named in this piece — Meta's own post of July 7, 2026 — is used to date the model, not to report an event.

What Meta Muse Video is, starting with the part that governs everything else

Meta Muse Video is an early preview, not a generally available model. That distinction is not a hedge; it is the fact the rest of the page follows from. Meta's own launch post, published under the byline Meta Superintelligence Labs on 2026-07-07 with the title "Introducing Muse Image and Muse Video", puts it plainly in its opening paragraph: "We're excited to launch Muse Image and preview Muse Video, the first media generation models developed by Meta Superintelligence Labs." One of the two shipped. The other was previewed. The availability line for the video model, repeated twice on the same page, is "Muse Video is coming soon to creators and Meta AI."

What that means for the Muse family as a whole is worth stating once, because it is easy to blur: Meta Muse Image and Meta Muse Video were introduced together and share a foundation — Meta says the video model is "built on the same pretraining base" as the image model — but they are at different stages. Meta Muse Image is a shipped product. Meta Muse Video is a demonstration with a name.

What Meta claims for it, and what Meta itself says is still weak

Meta's claims for Meta Muse Video are short, specific and vendor-reported, which is to say they come from Meta's own post and have not been independently reproduced by anyone we can cite:

• Competitive performance in prompt adherence, visual fidelity and temporal consistency. That is the claim Meta makes in its preview section, word for word, and it is a relative claim with no named comparison set attached to it.

• Native audio support. Meta's introduction describes Muse Video as delivering "exceptional visual fidelity with native audio support" — sound generated alongside picture rather than added afterwards. Meta Muse Image produces stills and has no audio surface, so this is the capability that separates the two.

• Two acknowledged gaps. In the same paragraph as the claim, Meta names what is not there yet: "We're investing in areas with current performance gaps, such as audio-video synchronization and physically accurate fast motion." Read that literally. A preview with native audio and imperfect audio-video synchronization is not a preview with reliable lip sync, and fast motion is called out as the other known weakness.

• A watermarking commitment, not a documented pipeline. Meta's invisible provenance system, Content Seal, ships with Muse Image; the post says Meta plans "to extend Content Seal to video soon". For a video-generation model that is the provenance answer as it stands today — planned, not described.

There is no benchmark in this set. Meta publishes no independent evaluation of Meta Muse Video's prompt adherence, fidelity or synchronization, and the numbers that do exist on Meta's page are internal ablations of the image model. An absent benchmark is a fact about the documentation, and it is the fact here.

Where Muse Video sits on Arena, and why you should not treat the number as stable

Meta states the model's leaderboard position in its own words: "On Arena, Muse Video ranks No. 3 in human-preference Elo for text-to-video at the time of writing," with the chart on that page captioned "Arena Elo rankings as of July 5, 2026". Treat all three qualifiers as load-bearing. It is a vendor-reported figure, it is human-preference Elo rather than a benchmark score, and it is explicitly bounded to a moment — the time of writing, on a chart dated two days before the post went up.

If you have seen a different number elsewhere, the honest description is that the position has moved over time. Human-preference boards take votes continuously and re-rank as field entries are added, so a rank captured in July is not the rank a board shows in September, and we are not going to reconcile those snapshots here: our existing page on the Muse Video leak holds the dated captures and the vote counts behind them, and it is the right place for them. Link the two pages in your head as one is to one: that page is the timeline, this page is the model. One caution that applies to any number you read anywhere, including Meta's — a board that scores a model with no public access is not measuring whether you can build on it.

What Muse Video is not

Three confusions are worth clearing, because they are the reason the searches behind this page exist:

• Not a shipped product inside Meta AI. The framing "coming soon to creators and Meta AI" is the current state of the release. There is nothing to open.

• Not on Meta's Model API. Meta's developer documentation introduces Model API with the models it serves — Muse Spark, Muse Image, Muse Voice Transcribe, Muse Glimmer and Segment Anything — and Meta Muse Video is not among them. The developer URL for it resolves to Meta's own "This page isn't available" screen. So the preview did not arrive with an endpoint, an identifier, a schema or a rate limit, and anyone telling you the API path exists is guessing.

• Not the same thing as Meta's realtime avatar work. Meta's research blog lists "Bringing Your Muse to Life" on 2026-09-23, which introduces Muse Realtime Avatar — an embodiment layer that turns Muse Realtime Voice into an interactive avatar. That is a live, conversational surface, not a clip generator, and it is a different product with a different reference page in our set. If what you want is generated video you can download and cut, neither of those is it either.

Meta Muse Image is the sibling you can actually use today

This is the practical answer for most readers who arrived searching for Meta Muse Video. Meta's image model is the one that shipped, and it is the Muse generation model you can use now.

• Where it runs (vendor-reported). Meta's post states Meta Muse Image is "available today across the Meta AI app and on meta.ai, Instagram Stories in the US, and WhatsApp in limited countries, and is coming soon to Facebook."

• What it costs where Meta publishes a price. Meta's own model page for Meta Muse Image states a flat "$0.01/image" and lists it with reasoning and tool use enabled, describing the model as "agentic image generation, grounded by search and priced for production volumes at $0.01/image". That is Meta's figure on Meta's page, and it is the only readable Muse Image price we can point at. Meta's rate card is not published in a form we could fetch: the pricing URLs under Meta's developer domain return a shell with no figures in it, so a price for any Muse model is reportable only when it is read on a model page, as this one was.

• What it does that a video model would not. Meta describes Meta Muse Image as an agentic generator — it plans, uses search and code tools, composes from multiple references, and refines its own work across turns. None of that is documented for Meta Muse Video.

There is no equivalent price, tier or limit to report for Meta Muse Video, because Meta has published none. We are not going to fill that gap from a sibling page, and neither should you.

The launch-date question, answered directly

No general availability date has been announced by Meta. There is no day, no month, no window and no commitment beyond "coming soon" on Meta's own pages as of 2026-09-29. You should read every specific date you see for Meta Muse Video as speculation unless it traces to a newer Meta announcement than the 2026-07-07 post — that includes leaked windows, calendar guesses and anything derived from a developer-platform trace. Our own existing coverage of the leak carries that constraint as its own rule and we are inheriting it verbatim here: a page quoting a specific launch day or a price for Meta Muse Video without a newer Meta announcement is speculating.

What is knowable is the shape of the open questions, and each one is a thing Meta has not addressed: the access path (consumer app first, developer endpoint, or both), clip length and resolution, whether native audio output is included in the first release or matures later, regional rollout, and commercial terms. If you are deciding whether Meta Muse Video changes a production plan this quarter, the answer today is that it cannot, because there is nothing to plan against.

Can you call Meta Muse Video through OrcaRouter? No — and here is what you can call

We checked our own live model list on 2026-09-29 rather than assuming, and Meta Muse Video is not served by us. That is the honest sentence, and it is the one a reader should expect for a model its own vendor has not released: a previewed model cannot be routable anywhere, and a gateway that told you otherwise would be wrong. Nothing about our Muse coverage should be read as implying otherwise — carrying Muse Spark does not mean carrying a Meta model Meta has not shipped.

What is real today is narrower and useful: Meta Muse Spark 1.2 is routable on OrcaRouter today, at Meta's own list price with no markup added by us, so a pipeline that wants Meta's reasoning model in it can have one now without waiting on a video endpoint that does not exist. And when Meta Muse Video does open — in whatever form — the same pattern applies to models we do serve: one key and one endpoint for the models we do route, a 0% markup pass-through that makes a vendor price change live on our side the same day it takes effect, and automatic failover so a model you are testing against a production path never becomes a single point of failure on its own. That is what a routing layer is for, and it is the only way this page can help you with a model you cannot call.

What to watch, and what would change this page

Three things would move Meta Muse Video from a preview to a subject you can act on, and each is a specific, checkable event rather than a rumour: an entry for it in Meta's developer model list, which is where an API surface would have to appear; a change to the availability line on Meta's own post or a new post under the Meta Superintelligence Labs byline; and a published price, which has not existed for this model in any form. Until one of those lands, this page's answer stands, and the most useful thing a reader can do with the wait is decide what would count as good enough. Write the brief you would test it against now — one motion-heavy shot, one piece of narration you would check for synchronization, one clip that has to hold a face together for its full length — and you will be able to judge the model on your material the day access opens instead of judging it on a demo reel.

That is the honest state of Meta Muse Video: a real model from a real lab, previewed rather than shipped, first in a pair of Meta Superintelligence Labs media generation models, ranked highly on a board at a moment in July that has since moved, and still without a date you can plan around.