Meta is running its unreleased Muse Video model through a closed beta with a small group of outside partners. TestingCatalog, a publication that tracks early access to unreleased AI products, says it obtained hands-on access and generated its own clips. The outlet’s samples show sharp fine detail, coherent scene logic, and motion that holds steady across a full ten-second clip, the model’s current output ceiling. Meta also confirmed the model generates native audio synced to the picture, though the company had previously flagged gaps in matching audio to fast motion.
The capability lands well after two rivals already ship native audio generation to the public. Google’s Veo 3 produces synced audio through the Gemini app and Vertex AI, and OpenAI’s Sora has been generally available for months. Meta unveiled Muse Video in July as a companion to its Muse Image model, saying then that video generation would roll out to creators and to Meta AI. A month later, the tool is still limited to select partners, with no confirmed date for a wider release.
The “state-of-the-art” read on Muse Video’s early output comes from TestingCatalog’s own hands-on testing, not from a published benchmark or a third-party evaluation. Meta has not released comparative scores against Veo 3, Sora, or Runway’s Gen-4 models. That gap matters: anyone weighing the claim is relying on one outlet’s judgment of a handful of clips rather than a standardized test anyone else can rerun.
Meta’s actual advantage sits in distribution, not necessarily quality. Muse Image already runs free inside Meta AI, and the company has said Muse Video is headed to the same surface. From there, the model could feed Meta’s Vibes feed, the short-form AI video stream built into its apps, and extend into Instagram, Facebook, and the standalone Edits creation app. No other video lab can put a new model in front of billions of existing Instagram and Facebook users on day one. That reach turns a beta test into a real threat to standalone generators, regardless of how Muse Video ranks on a leaderboard.
Pricing and generation limits for Muse Video remain unannounced. Muse Image’s free-to-use structure raises the question of whether Meta subsidizes video generation too, or caps it once usage moves past a small testing group. Ten-second clips, if they ship free at consumer scale, would pressure paid generators that meter output by the second and charge accordingly.
Two technical gaps remain open. Meta has acknowledged that audio-video synchronization is not fully solved, and that fast motion is not always rendered with physically accurate results. Neither issue is disqualifying at the beta stage, but both will determine whether Muse Video is usable for anything beyond short, low-motion social clips at launch.
For product teams evaluating video-generation vendors, Muse Video is not yet buyable at any price. Teams that need production-ready generative video now should keep building on Veo 3 or Sora, and revisit Meta’s stack once a public release date, the audio-sync fix, and pricing are confirmed.
TestingCatalog reported the hands-on access and published early Muse Video samples on August 19, 2026.
