Wan2.2 video generation models — T2V/I2V MoE (14B active) and unified TI2V (5B dense), served via vLLM-Omni
View on HuggingFace
View on ModelScope