China has unveiled Wan2.2, an advanced open-source AI video model suite by Alibaba’s Tongyi Lab. Built on a Mixture-of-Experts (MoE) framework, it offers text-to-video (T2V-A14B), image-to-video (I2V-A14B), and a hybrid TI2V-5B model for both. Wan2.2 enables cinematic precision—controlling lighting, tone, and camera angles—while improving motion realism in facial expressions, gestures, and dynamic scenes. Its dual-expert denoising cuts computation by about 50%, making high-quality results faster and more efficient. The hybrid 5B model even runs on consumer GPUs, generating smooth 720p videos. This breakthrough makes cinematic AI video creation more accessible to creators worldwide. #Wan2point2 #AlibabaAI #VideoAI #CinematicAI #AIInnovation #OpenSource #DigitalCreativity #AItools