AnimateAnything is a unified controllable video generation method that supports precise and consistent video operations under various conditions, including camera trajectories, text prompts, and user action annotations. This technology constructs a universal motion representation under different conditions through a multi-scale feature fusion network, converting all control information into frame-by-frame optical flows to guide video generation. Furthermore, to mitigate flickering issues caused by large-scale motion, a frequency-based stabilization module is proposed to ensure consistency in the frequency domain and enhance temporal coherence. Experimental results demonstrate that the AnimateAnything method outperforms existing state-of-the-art techniques.