Meta Unveils Muse Video Model: 10-Second AI Clips in Beta

On August 20, according to IT Home, tech media outlet testingcatalog published a post yesterday reporting that Meta is set to launch its Muse Video model, which has already invited select partners for beta testing, capable of generating up to 10-second videos in a single run. Based on videos generated during testing, Muse Video demonstrates high potential in detail rendering, understanding of object relationships in scenes, and temporal stability across frames.

Meta has previously acknowledged that the model still faces issues with audio-video synchronization and physical accuracy for high-speed motion. Current samples are limited to 10-second clips, which are insufficient to validate performance in longer narrative sequences, sustained character consistency, or high-intensity action scenes.