Wan-AI/Wan2.2-I2V-A14B

Wan2.2-I2V-A14B is one of the industry's first open-source image-to-video generation models to adopt a Mixture-of-Experts (MoE) architecture, released by Wan-AI, an AI project under Alibaba. This model focuses on combining static images with text prompts to generate smooth, natural dynamic video sequences. Its core innovation lies in the MoE architecture, which employs a high-noise expert to handle the overall structure of early-stage videos and a low-noise expert to refine detailed visual elements in later stages—thereby enhancing model performance without increasing inference costs. Compared to its predecessor, Wan2.2 was trained on a much larger dataset, significantly improving its ability to understand complex motions, aesthetic styles, and semantics, resulting in more stable video outputs and reduced instances of unrealistic camera movements.