According to DeepLearning.AI, Alibaba has released Wan 2.2, a new open-weights video generation AI model family featuring an advanced mixture-of-experts (MoE) architecture. The Wan 2.2 family includes a five-billion-parameter text/image-to-video model that is optimized for consumer GPUs, lowering the barrier for small businesses and developers to access high-performance video synthesis. The MoE design employs two specialized experts—one for handling high-noise data and another for low-noise scenarios—improving video generation accuracy and efficiency. This development signals increased accessibility and scalability in AI-powered video content creation, opening new opportunities for content creators, marketers, and AI startups to leverage advanced generative video tools without the need for expensive hardware (source: DeepLearning.AI).