🌀 Meta Muse Video: What We Know & Why It Matters

Meta finally entered the AI video race on July 7 — and managed to launch a genuinely strong model and a privacy firestorm in the same week. Muse Video is the first video model from Meta Superintelligence Labs, previewed alongside the Muse Image generator, and the early leaderboard numbers say it belongs in the conversation with the models in our tested ranking. Here's what's confirmed, what isn't, and what the launch mess tells creators.

By the numbers

• Previewed July 7, 2026 by Meta Superintelligence Labs, alongside the public Muse Image launch (ForkLog)
• Debuted at #3 on the Arena text-to-video leaderboard — an unusually strong first showing (Pexo)
• Native audio generation — joining Veo and Sora in the sound-included club, ahead of most rivals
• 3 days — how long Muse Image's @-mention face-generation feature survived before Meta pulled it over consent backlash (Republic World)
• Availability: preview-only — no public resolution, duration, or pricing specs yet

What's actually confirmed

Three things: it exists, it generates video from prompts with native audio on the Muse Image pretraining base, and third-party arena voters rank it top-3 for text-to-video. That last point deserves respect — models rarely debut that high. What's not public: resolution ceilings, clip length, image-to-video support, reference/consistency tooling, moderation behavior, API plans, or price. In other words: everything a production workflow actually depends on.

The face-cloning mess, and why it matters to creators

Muse Image shipped with a feature that let anyone @-mention a public Instagram account and generate synthetic images of that person — no notification, no opt-in. The backlash was immediate; Meta killed the feature in three days. Two takeaways for working creators: first, expect Muse Video's likeness rules to launch strict — Meta just learned this lesson publicly and expensively. Second, if your content involves real people, consent documentation is becoming the industry's default expectation, not a nicety. (Your own face is your call — check what Meta's settings expose about your public photos either way.)

Where Muse Video fits — if the specs land

Meta's distribution is the story. Every other frontier video model makes you come to it; Muse ships inside Meta AI, Instagram and Facebook surfaces the moment it's public. That's a different game: less "production tool," more "the default video generator for a billion casual users" — the same split we describe between app-first Sora and production-first Seedance. Serious pipelines care about controllability (references, multi-shot grammar, predictable moderation), and none of that is known yet.

Our verdict policy

We don't rank previews. Muse Video enters the best AI video generators ranking when we can run identical prompts on it and its rivals, measure cost per finished shot, and map its moderation quirks — the same bar every model on this site clears. This page updates as specs land (the updated date above is honest).

What to do today

Nothing changes for production this week: Seedance and Veo remain the workhorses, and Sora 2 remains the social-native option. If you want to be early on Muse, the practical move is watching the Meta AI surfaces where it will land (meta.ai) — and getting your prompting fundamentals sharp on current models, because those skills transfer to every new launch. When Muse Video opens up, we'll test it the same week and publish the numbers.